Merge tag 'for-5.3/libata-20190708' of git://git.kernel.dk/linux-block
[linux-2.6-block.git] / fs / nfs / nfs4proc.c
1 /*
2  *  fs/nfs/nfs4proc.c
3  *
4  *  Client-side procedure declarations for NFSv4.
5  *
6  *  Copyright (c) 2002 The Regents of the University of Michigan.
7  *  All rights reserved.
8  *
9  *  Kendrick Smith <kmsmith@umich.edu>
10  *  Andy Adamson   <andros@umich.edu>
11  *
12  *  Redistribution and use in source and binary forms, with or without
13  *  modification, are permitted provided that the following conditions
14  *  are met:
15  *
16  *  1. Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  *  2. Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  *  3. Neither the name of the University nor the names of its
22  *     contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36  */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/xattr.h>
55 #include <linux/utsname.h>
56 #include <linux/freezer.h>
57 #include <linux/iversion.h>
58
59 #include "nfs4_fs.h"
60 #include "delegation.h"
61 #include "internal.h"
62 #include "iostat.h"
63 #include "callback.h"
64 #include "pnfs.h"
65 #include "netns.h"
66 #include "nfs4idmap.h"
67 #include "nfs4session.h"
68 #include "fscache.h"
69
70 #include "nfs4trace.h"
71
72 #define NFSDBG_FACILITY         NFSDBG_PROC
73
74 #define NFS4_BITMASK_SZ         3
75
76 #define NFS4_POLL_RETRY_MIN     (HZ/10)
77 #define NFS4_POLL_RETRY_MAX     (15*HZ)
78
79 /* file attributes which can be mapped to nfs attributes */
80 #define NFS4_VALID_ATTRS (ATTR_MODE \
81         | ATTR_UID \
82         | ATTR_GID \
83         | ATTR_SIZE \
84         | ATTR_ATIME \
85         | ATTR_MTIME \
86         | ATTR_CTIME \
87         | ATTR_ATIME_SET \
88         | ATTR_MTIME_SET)
89
90 struct nfs4_opendata;
91 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
92 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
93 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
94 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label, struct inode *inode);
95 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
96 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
97                             struct nfs_fattr *fattr, struct iattr *sattr,
98                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
99                             struct nfs4_label *olabel);
100 #ifdef CONFIG_NFS_V4_1
101 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
102                 const struct cred *cred,
103                 struct nfs4_slot *slot,
104                 bool is_privileged);
105 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
106                 const struct cred *);
107 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
108                 const struct cred *, bool);
109 #endif
110
111 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
112 static inline struct nfs4_label *
113 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
114         struct iattr *sattr, struct nfs4_label *label)
115 {
116         int err;
117
118         if (label == NULL)
119                 return NULL;
120
121         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
122                 return NULL;
123
124         err = security_dentry_init_security(dentry, sattr->ia_mode,
125                                 &dentry->d_name, (void **)&label->label, &label->len);
126         if (err == 0)
127                 return label;
128
129         return NULL;
130 }
131 static inline void
132 nfs4_label_release_security(struct nfs4_label *label)
133 {
134         if (label)
135                 security_release_secctx(label->label, label->len);
136 }
137 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
138 {
139         if (label)
140                 return server->attr_bitmask;
141
142         return server->attr_bitmask_nl;
143 }
144 #else
145 static inline struct nfs4_label *
146 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
147         struct iattr *sattr, struct nfs4_label *l)
148 { return NULL; }
149 static inline void
150 nfs4_label_release_security(struct nfs4_label *label)
151 { return; }
152 static inline u32 *
153 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
154 { return server->attr_bitmask; }
155 #endif
156
157 /* Prevent leaks of NFSv4 errors into userland */
158 static int nfs4_map_errors(int err)
159 {
160         if (err >= -1000)
161                 return err;
162         switch (err) {
163         case -NFS4ERR_RESOURCE:
164         case -NFS4ERR_LAYOUTTRYLATER:
165         case -NFS4ERR_RECALLCONFLICT:
166                 return -EREMOTEIO;
167         case -NFS4ERR_WRONGSEC:
168         case -NFS4ERR_WRONG_CRED:
169                 return -EPERM;
170         case -NFS4ERR_BADOWNER:
171         case -NFS4ERR_BADNAME:
172                 return -EINVAL;
173         case -NFS4ERR_SHARE_DENIED:
174                 return -EACCES;
175         case -NFS4ERR_MINOR_VERS_MISMATCH:
176                 return -EPROTONOSUPPORT;
177         case -NFS4ERR_FILE_OPEN:
178                 return -EBUSY;
179         default:
180                 dprintk("%s could not handle NFSv4 error %d\n",
181                                 __func__, -err);
182                 break;
183         }
184         return -EIO;
185 }
186
187 /*
188  * This is our standard bitmap for GETATTR requests.
189  */
190 const u32 nfs4_fattr_bitmap[3] = {
191         FATTR4_WORD0_TYPE
192         | FATTR4_WORD0_CHANGE
193         | FATTR4_WORD0_SIZE
194         | FATTR4_WORD0_FSID
195         | FATTR4_WORD0_FILEID,
196         FATTR4_WORD1_MODE
197         | FATTR4_WORD1_NUMLINKS
198         | FATTR4_WORD1_OWNER
199         | FATTR4_WORD1_OWNER_GROUP
200         | FATTR4_WORD1_RAWDEV
201         | FATTR4_WORD1_SPACE_USED
202         | FATTR4_WORD1_TIME_ACCESS
203         | FATTR4_WORD1_TIME_METADATA
204         | FATTR4_WORD1_TIME_MODIFY
205         | FATTR4_WORD1_MOUNTED_ON_FILEID,
206 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
207         FATTR4_WORD2_SECURITY_LABEL
208 #endif
209 };
210
211 static const u32 nfs4_pnfs_open_bitmap[3] = {
212         FATTR4_WORD0_TYPE
213         | FATTR4_WORD0_CHANGE
214         | FATTR4_WORD0_SIZE
215         | FATTR4_WORD0_FSID
216         | FATTR4_WORD0_FILEID,
217         FATTR4_WORD1_MODE
218         | FATTR4_WORD1_NUMLINKS
219         | FATTR4_WORD1_OWNER
220         | FATTR4_WORD1_OWNER_GROUP
221         | FATTR4_WORD1_RAWDEV
222         | FATTR4_WORD1_SPACE_USED
223         | FATTR4_WORD1_TIME_ACCESS
224         | FATTR4_WORD1_TIME_METADATA
225         | FATTR4_WORD1_TIME_MODIFY,
226         FATTR4_WORD2_MDSTHRESHOLD
227 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
228         | FATTR4_WORD2_SECURITY_LABEL
229 #endif
230 };
231
232 static const u32 nfs4_open_noattr_bitmap[3] = {
233         FATTR4_WORD0_TYPE
234         | FATTR4_WORD0_FILEID,
235 };
236
237 const u32 nfs4_statfs_bitmap[3] = {
238         FATTR4_WORD0_FILES_AVAIL
239         | FATTR4_WORD0_FILES_FREE
240         | FATTR4_WORD0_FILES_TOTAL,
241         FATTR4_WORD1_SPACE_AVAIL
242         | FATTR4_WORD1_SPACE_FREE
243         | FATTR4_WORD1_SPACE_TOTAL
244 };
245
246 const u32 nfs4_pathconf_bitmap[3] = {
247         FATTR4_WORD0_MAXLINK
248         | FATTR4_WORD0_MAXNAME,
249         0
250 };
251
252 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
253                         | FATTR4_WORD0_MAXREAD
254                         | FATTR4_WORD0_MAXWRITE
255                         | FATTR4_WORD0_LEASE_TIME,
256                         FATTR4_WORD1_TIME_DELTA
257                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
258                         FATTR4_WORD2_LAYOUT_BLKSIZE
259                         | FATTR4_WORD2_CLONE_BLKSIZE
260 };
261
262 const u32 nfs4_fs_locations_bitmap[3] = {
263         FATTR4_WORD0_CHANGE
264         | FATTR4_WORD0_SIZE
265         | FATTR4_WORD0_FSID
266         | FATTR4_WORD0_FILEID
267         | FATTR4_WORD0_FS_LOCATIONS,
268         FATTR4_WORD1_OWNER
269         | FATTR4_WORD1_OWNER_GROUP
270         | FATTR4_WORD1_RAWDEV
271         | FATTR4_WORD1_SPACE_USED
272         | FATTR4_WORD1_TIME_ACCESS
273         | FATTR4_WORD1_TIME_METADATA
274         | FATTR4_WORD1_TIME_MODIFY
275         | FATTR4_WORD1_MOUNTED_ON_FILEID,
276 };
277
278 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
279                 struct inode *inode)
280 {
281         unsigned long cache_validity;
282
283         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
284         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
285                 return;
286
287         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
288         if (!(cache_validity & NFS_INO_REVAL_FORCED))
289                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
290                                 | NFS_INO_INVALID_SIZE);
291
292         if (!(cache_validity & NFS_INO_INVALID_SIZE))
293                 dst[0] &= ~FATTR4_WORD0_SIZE;
294
295         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
296                 dst[0] &= ~FATTR4_WORD0_CHANGE;
297 }
298
299 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
300                 const __u32 *src, struct inode *inode)
301 {
302         nfs4_bitmap_copy_adjust(dst, src, inode);
303 }
304
305 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
306                 struct nfs4_readdir_arg *readdir)
307 {
308         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
309         __be32 *start, *p;
310
311         if (cookie > 2) {
312                 readdir->cookie = cookie;
313                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
314                 return;
315         }
316
317         readdir->cookie = 0;
318         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
319         if (cookie == 2)
320                 return;
321         
322         /*
323          * NFSv4 servers do not return entries for '.' and '..'
324          * Therefore, we fake these entries here.  We let '.'
325          * have cookie 0 and '..' have cookie 1.  Note that
326          * when talking to the server, we always send cookie 0
327          * instead of 1 or 2.
328          */
329         start = p = kmap_atomic(*readdir->pages);
330         
331         if (cookie == 0) {
332                 *p++ = xdr_one;                                  /* next */
333                 *p++ = xdr_zero;                   /* cookie, first word */
334                 *p++ = xdr_one;                   /* cookie, second word */
335                 *p++ = xdr_one;                             /* entry len */
336                 memcpy(p, ".\0\0\0", 4);                        /* entry */
337                 p++;
338                 *p++ = xdr_one;                         /* bitmap length */
339                 *p++ = htonl(attrs);                           /* bitmap */
340                 *p++ = htonl(12);             /* attribute buffer length */
341                 *p++ = htonl(NF4DIR);
342                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
343         }
344         
345         *p++ = xdr_one;                                  /* next */
346         *p++ = xdr_zero;                   /* cookie, first word */
347         *p++ = xdr_two;                   /* cookie, second word */
348         *p++ = xdr_two;                             /* entry len */
349         memcpy(p, "..\0\0", 4);                         /* entry */
350         p++;
351         *p++ = xdr_one;                         /* bitmap length */
352         *p++ = htonl(attrs);                           /* bitmap */
353         *p++ = htonl(12);             /* attribute buffer length */
354         *p++ = htonl(NF4DIR);
355         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
356
357         readdir->pgbase = (char *)p - (char *)start;
358         readdir->count -= readdir->pgbase;
359         kunmap_atomic(start);
360 }
361
362 static void nfs4_test_and_free_stateid(struct nfs_server *server,
363                 nfs4_stateid *stateid,
364                 const struct cred *cred)
365 {
366         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
367
368         ops->test_and_free_expired(server, stateid, cred);
369 }
370
371 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
372                 nfs4_stateid *stateid,
373                 const struct cred *cred)
374 {
375         stateid->type = NFS4_REVOKED_STATEID_TYPE;
376         nfs4_test_and_free_stateid(server, stateid, cred);
377 }
378
379 static void nfs4_free_revoked_stateid(struct nfs_server *server,
380                 const nfs4_stateid *stateid,
381                 const struct cred *cred)
382 {
383         nfs4_stateid tmp;
384
385         nfs4_stateid_copy(&tmp, stateid);
386         __nfs4_free_revoked_stateid(server, &tmp, cred);
387 }
388
389 static long nfs4_update_delay(long *timeout)
390 {
391         long ret;
392         if (!timeout)
393                 return NFS4_POLL_RETRY_MAX;
394         if (*timeout <= 0)
395                 *timeout = NFS4_POLL_RETRY_MIN;
396         if (*timeout > NFS4_POLL_RETRY_MAX)
397                 *timeout = NFS4_POLL_RETRY_MAX;
398         ret = *timeout;
399         *timeout <<= 1;
400         return ret;
401 }
402
403 static int nfs4_delay_killable(long *timeout)
404 {
405         might_sleep();
406
407         freezable_schedule_timeout_killable_unsafe(
408                 nfs4_update_delay(timeout));
409         if (!__fatal_signal_pending(current))
410                 return 0;
411         return -EINTR;
412 }
413
414 static int nfs4_delay_interruptible(long *timeout)
415 {
416         might_sleep();
417
418         freezable_schedule_timeout_interruptible(nfs4_update_delay(timeout));
419         if (!signal_pending(current))
420                 return 0;
421         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
422 }
423
424 static int nfs4_delay(long *timeout, bool interruptible)
425 {
426         if (interruptible)
427                 return nfs4_delay_interruptible(timeout);
428         return nfs4_delay_killable(timeout);
429 }
430
431 /* This is the error handling routine for processes that are allowed
432  * to sleep.
433  */
434 static int nfs4_do_handle_exception(struct nfs_server *server,
435                 int errorcode, struct nfs4_exception *exception)
436 {
437         struct nfs_client *clp = server->nfs_client;
438         struct nfs4_state *state = exception->state;
439         const nfs4_stateid *stateid = exception->stateid;
440         struct inode *inode = exception->inode;
441         int ret = errorcode;
442
443         exception->delay = 0;
444         exception->recovering = 0;
445         exception->retry = 0;
446
447         if (stateid == NULL && state != NULL)
448                 stateid = &state->stateid;
449
450         switch(errorcode) {
451                 case 0:
452                         return 0;
453                 case -NFS4ERR_BADHANDLE:
454                 case -ESTALE:
455                         if (inode != NULL && S_ISREG(inode->i_mode))
456                                 pnfs_destroy_layout(NFS_I(inode));
457                         break;
458                 case -NFS4ERR_DELEG_REVOKED:
459                 case -NFS4ERR_ADMIN_REVOKED:
460                 case -NFS4ERR_EXPIRED:
461                 case -NFS4ERR_BAD_STATEID:
462                         if (inode != NULL && stateid != NULL) {
463                                 nfs_inode_find_state_and_recover(inode,
464                                                 stateid);
465                                 goto wait_on_recovery;
466                         }
467                         /* Fall through */
468                 case -NFS4ERR_OPENMODE:
469                         if (inode) {
470                                 int err;
471
472                                 err = nfs_async_inode_return_delegation(inode,
473                                                 stateid);
474                                 if (err == 0)
475                                         goto wait_on_recovery;
476                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
477                                         exception->retry = 1;
478                                         break;
479                                 }
480                         }
481                         if (state == NULL)
482                                 break;
483                         ret = nfs4_schedule_stateid_recovery(server, state);
484                         if (ret < 0)
485                                 break;
486                         goto wait_on_recovery;
487                 case -NFS4ERR_STALE_STATEID:
488                 case -NFS4ERR_STALE_CLIENTID:
489                         nfs4_schedule_lease_recovery(clp);
490                         goto wait_on_recovery;
491                 case -NFS4ERR_MOVED:
492                         ret = nfs4_schedule_migration_recovery(server);
493                         if (ret < 0)
494                                 break;
495                         goto wait_on_recovery;
496                 case -NFS4ERR_LEASE_MOVED:
497                         nfs4_schedule_lease_moved_recovery(clp);
498                         goto wait_on_recovery;
499 #if defined(CONFIG_NFS_V4_1)
500                 case -NFS4ERR_BADSESSION:
501                 case -NFS4ERR_BADSLOT:
502                 case -NFS4ERR_BAD_HIGH_SLOT:
503                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
504                 case -NFS4ERR_DEADSESSION:
505                 case -NFS4ERR_SEQ_FALSE_RETRY:
506                 case -NFS4ERR_SEQ_MISORDERED:
507                         dprintk("%s ERROR: %d Reset session\n", __func__,
508                                 errorcode);
509                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
510                         goto wait_on_recovery;
511 #endif /* defined(CONFIG_NFS_V4_1) */
512                 case -NFS4ERR_FILE_OPEN:
513                         if (exception->timeout > HZ) {
514                                 /* We have retried a decent amount, time to
515                                  * fail
516                                  */
517                                 ret = -EBUSY;
518                                 break;
519                         }
520                         /* Fall through */
521                 case -NFS4ERR_DELAY:
522                         nfs_inc_server_stats(server, NFSIOS_DELAY);
523                         /* Fall through */
524                 case -NFS4ERR_GRACE:
525                 case -NFS4ERR_LAYOUTTRYLATER:
526                 case -NFS4ERR_RECALLCONFLICT:
527                         exception->delay = 1;
528                         return 0;
529
530                 case -NFS4ERR_RETRY_UNCACHED_REP:
531                 case -NFS4ERR_OLD_STATEID:
532                         exception->retry = 1;
533                         break;
534                 case -NFS4ERR_BADOWNER:
535                         /* The following works around a Linux server bug! */
536                 case -NFS4ERR_BADNAME:
537                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
538                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
539                                 exception->retry = 1;
540                                 printk(KERN_WARNING "NFS: v4 server %s "
541                                                 "does not accept raw "
542                                                 "uid/gids. "
543                                                 "Reenabling the idmapper.\n",
544                                                 server->nfs_client->cl_hostname);
545                         }
546         }
547         /* We failed to handle the error */
548         return nfs4_map_errors(ret);
549 wait_on_recovery:
550         exception->recovering = 1;
551         return 0;
552 }
553
554 /* This is the error handling routine for processes that are allowed
555  * to sleep.
556  */
557 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
558 {
559         struct nfs_client *clp = server->nfs_client;
560         int ret;
561
562         ret = nfs4_do_handle_exception(server, errorcode, exception);
563         if (exception->delay) {
564                 ret = nfs4_delay(&exception->timeout,
565                                 exception->interruptible);
566                 goto out_retry;
567         }
568         if (exception->recovering) {
569                 ret = nfs4_wait_clnt_recover(clp);
570                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
571                         return -EIO;
572                 goto out_retry;
573         }
574         return ret;
575 out_retry:
576         if (ret == 0)
577                 exception->retry = 1;
578         return ret;
579 }
580
581 static int
582 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
583                 int errorcode, struct nfs4_exception *exception)
584 {
585         struct nfs_client *clp = server->nfs_client;
586         int ret;
587
588         ret = nfs4_do_handle_exception(server, errorcode, exception);
589         if (exception->delay) {
590                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
591                 goto out_retry;
592         }
593         if (exception->recovering) {
594                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
595                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
596                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
597                 goto out_retry;
598         }
599         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
600                 ret = -EIO;
601         return ret;
602 out_retry:
603         if (ret == 0) {
604                 exception->retry = 1;
605                 /*
606                  * For NFS4ERR_MOVED, the client transport will need to
607                  * be recomputed after migration recovery has completed.
608                  */
609                 if (errorcode == -NFS4ERR_MOVED)
610                         rpc_task_release_transport(task);
611         }
612         return ret;
613 }
614
615 int
616 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
617                         struct nfs4_state *state, long *timeout)
618 {
619         struct nfs4_exception exception = {
620                 .state = state,
621         };
622
623         if (task->tk_status >= 0)
624                 return 0;
625         if (timeout)
626                 exception.timeout = *timeout;
627         task->tk_status = nfs4_async_handle_exception(task, server,
628                         task->tk_status,
629                         &exception);
630         if (exception.delay && timeout)
631                 *timeout = exception.timeout;
632         if (exception.retry)
633                 return -EAGAIN;
634         return 0;
635 }
636
637 /*
638  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
639  * or 'false' otherwise.
640  */
641 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
642 {
643         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
644         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
645 }
646
647 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
648 {
649         spin_lock(&clp->cl_lock);
650         if (time_before(clp->cl_last_renewal,timestamp))
651                 clp->cl_last_renewal = timestamp;
652         spin_unlock(&clp->cl_lock);
653 }
654
655 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
656 {
657         struct nfs_client *clp = server->nfs_client;
658
659         if (!nfs4_has_session(clp))
660                 do_renew_lease(clp, timestamp);
661 }
662
663 struct nfs4_call_sync_data {
664         const struct nfs_server *seq_server;
665         struct nfs4_sequence_args *seq_args;
666         struct nfs4_sequence_res *seq_res;
667 };
668
669 void nfs4_init_sequence(struct nfs4_sequence_args *args,
670                         struct nfs4_sequence_res *res, int cache_reply,
671                         int privileged)
672 {
673         args->sa_slot = NULL;
674         args->sa_cache_this = cache_reply;
675         args->sa_privileged = privileged;
676
677         res->sr_slot = NULL;
678 }
679
680 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
681 {
682         struct nfs4_slot *slot = res->sr_slot;
683         struct nfs4_slot_table *tbl;
684
685         tbl = slot->table;
686         spin_lock(&tbl->slot_tbl_lock);
687         if (!nfs41_wake_and_assign_slot(tbl, slot))
688                 nfs4_free_slot(tbl, slot);
689         spin_unlock(&tbl->slot_tbl_lock);
690
691         res->sr_slot = NULL;
692 }
693
694 static int nfs40_sequence_done(struct rpc_task *task,
695                                struct nfs4_sequence_res *res)
696 {
697         if (res->sr_slot != NULL)
698                 nfs40_sequence_free_slot(res);
699         return 1;
700 }
701
702 #if defined(CONFIG_NFS_V4_1)
703
704 static void nfs41_release_slot(struct nfs4_slot *slot)
705 {
706         struct nfs4_session *session;
707         struct nfs4_slot_table *tbl;
708         bool send_new_highest_used_slotid = false;
709
710         if (!slot)
711                 return;
712         tbl = slot->table;
713         session = tbl->session;
714
715         /* Bump the slot sequence number */
716         if (slot->seq_done)
717                 slot->seq_nr++;
718         slot->seq_done = 0;
719
720         spin_lock(&tbl->slot_tbl_lock);
721         /* Be nice to the server: try to ensure that the last transmitted
722          * value for highest_user_slotid <= target_highest_slotid
723          */
724         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
725                 send_new_highest_used_slotid = true;
726
727         if (nfs41_wake_and_assign_slot(tbl, slot)) {
728                 send_new_highest_used_slotid = false;
729                 goto out_unlock;
730         }
731         nfs4_free_slot(tbl, slot);
732
733         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
734                 send_new_highest_used_slotid = false;
735 out_unlock:
736         spin_unlock(&tbl->slot_tbl_lock);
737         if (send_new_highest_used_slotid)
738                 nfs41_notify_server(session->clp);
739         if (waitqueue_active(&tbl->slot_waitq))
740                 wake_up_all(&tbl->slot_waitq);
741 }
742
743 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
744 {
745         nfs41_release_slot(res->sr_slot);
746         res->sr_slot = NULL;
747 }
748
749 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
750                 u32 seqnr)
751 {
752         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
753                 slot->seq_nr_highest_sent = seqnr;
754 }
755 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
756                 u32 seqnr)
757 {
758         slot->seq_nr_highest_sent = seqnr;
759         slot->seq_nr_last_acked = seqnr;
760 }
761
762 static int nfs41_sequence_process(struct rpc_task *task,
763                 struct nfs4_sequence_res *res)
764 {
765         struct nfs4_session *session;
766         struct nfs4_slot *slot = res->sr_slot;
767         struct nfs_client *clp;
768         int ret = 1;
769
770         if (slot == NULL)
771                 goto out_noaction;
772         /* don't increment the sequence number if the task wasn't sent */
773         if (!RPC_WAS_SENT(task) || slot->seq_done)
774                 goto out;
775
776         session = slot->table->session;
777
778         trace_nfs4_sequence_done(session, res);
779         /* Check the SEQUENCE operation status */
780         switch (res->sr_status) {
781         case 0:
782                 /* Mark this sequence number as having been acked */
783                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
784                 /* Update the slot's sequence and clientid lease timer */
785                 slot->seq_done = 1;
786                 clp = session->clp;
787                 do_renew_lease(clp, res->sr_timestamp);
788                 /* Check sequence flags */
789                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
790                                 !!slot->privileged);
791                 nfs41_update_target_slotid(slot->table, slot, res);
792                 break;
793         case 1:
794                 /*
795                  * sr_status remains 1 if an RPC level error occurred.
796                  * The server may or may not have processed the sequence
797                  * operation..
798                  */
799                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
800                 slot->seq_done = 1;
801                 goto out;
802         case -NFS4ERR_DELAY:
803                 /* The server detected a resend of the RPC call and
804                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
805                  * of RFC5661.
806                  */
807                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
808                         __func__,
809                         slot->slot_nr,
810                         slot->seq_nr);
811                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
812                 goto out_retry;
813         case -NFS4ERR_RETRY_UNCACHED_REP:
814         case -NFS4ERR_SEQ_FALSE_RETRY:
815                 /*
816                  * The server thinks we tried to replay a request.
817                  * Retry the call after bumping the sequence ID.
818                  */
819                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
820                 goto retry_new_seq;
821         case -NFS4ERR_BADSLOT:
822                 /*
823                  * The slot id we used was probably retired. Try again
824                  * using a different slot id.
825                  */
826                 if (slot->slot_nr < slot->table->target_highest_slotid)
827                         goto session_recover;
828                 goto retry_nowait;
829         case -NFS4ERR_SEQ_MISORDERED:
830                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
831                 /*
832                  * Were one or more calls using this slot interrupted?
833                  * If the server never received the request, then our
834                  * transmitted slot sequence number may be too high.
835                  */
836                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
837                         slot->seq_nr--;
838                         goto retry_nowait;
839                 }
840                 /*
841                  * RFC5661:
842                  * A retry might be sent while the original request is
843                  * still in progress on the replier. The replier SHOULD
844                  * deal with the issue by returning NFS4ERR_DELAY as the
845                  * reply to SEQUENCE or CB_SEQUENCE operation, but
846                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
847                  *
848                  * Restart the search after a delay.
849                  */
850                 slot->seq_nr = slot->seq_nr_highest_sent;
851                 goto out_retry;
852         default:
853                 /* Just update the slot sequence no. */
854                 slot->seq_done = 1;
855         }
856 out:
857         /* The session may be reset by one of the error handlers. */
858         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
859 out_noaction:
860         return ret;
861 session_recover:
862         nfs4_schedule_session_recovery(session, res->sr_status);
863         goto retry_nowait;
864 retry_new_seq:
865         ++slot->seq_nr;
866 retry_nowait:
867         if (rpc_restart_call_prepare(task)) {
868                 nfs41_sequence_free_slot(res);
869                 task->tk_status = 0;
870                 ret = 0;
871         }
872         goto out;
873 out_retry:
874         if (!rpc_restart_call(task))
875                 goto out;
876         rpc_delay(task, NFS4_POLL_RETRY_MAX);
877         return 0;
878 }
879
880 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
881 {
882         if (!nfs41_sequence_process(task, res))
883                 return 0;
884         if (res->sr_slot != NULL)
885                 nfs41_sequence_free_slot(res);
886         return 1;
887
888 }
889 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
890
891 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
892 {
893         if (res->sr_slot == NULL)
894                 return 1;
895         if (res->sr_slot->table->session != NULL)
896                 return nfs41_sequence_process(task, res);
897         return nfs40_sequence_done(task, res);
898 }
899
900 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
901 {
902         if (res->sr_slot != NULL) {
903                 if (res->sr_slot->table->session != NULL)
904                         nfs41_sequence_free_slot(res);
905                 else
906                         nfs40_sequence_free_slot(res);
907         }
908 }
909
910 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
911 {
912         if (res->sr_slot == NULL)
913                 return 1;
914         if (!res->sr_slot->table->session)
915                 return nfs40_sequence_done(task, res);
916         return nfs41_sequence_done(task, res);
917 }
918 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
919
920 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
921 {
922         struct nfs4_call_sync_data *data = calldata;
923
924         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
925
926         nfs4_setup_sequence(data->seq_server->nfs_client,
927                             data->seq_args, data->seq_res, task);
928 }
929
930 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
931 {
932         struct nfs4_call_sync_data *data = calldata;
933
934         nfs41_sequence_done(task, data->seq_res);
935 }
936
937 static const struct rpc_call_ops nfs41_call_sync_ops = {
938         .rpc_call_prepare = nfs41_call_sync_prepare,
939         .rpc_call_done = nfs41_call_sync_done,
940 };
941
942 #else   /* !CONFIG_NFS_V4_1 */
943
944 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
945 {
946         return nfs40_sequence_done(task, res);
947 }
948
949 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
950 {
951         if (res->sr_slot != NULL)
952                 nfs40_sequence_free_slot(res);
953 }
954
955 int nfs4_sequence_done(struct rpc_task *task,
956                        struct nfs4_sequence_res *res)
957 {
958         return nfs40_sequence_done(task, res);
959 }
960 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
961
962 #endif  /* !CONFIG_NFS_V4_1 */
963
964 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
965 {
966         res->sr_timestamp = jiffies;
967         res->sr_status_flags = 0;
968         res->sr_status = 1;
969 }
970
971 static
972 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
973                 struct nfs4_sequence_res *res,
974                 struct nfs4_slot *slot)
975 {
976         if (!slot)
977                 return;
978         slot->privileged = args->sa_privileged ? 1 : 0;
979         args->sa_slot = slot;
980
981         res->sr_slot = slot;
982 }
983
984 int nfs4_setup_sequence(struct nfs_client *client,
985                         struct nfs4_sequence_args *args,
986                         struct nfs4_sequence_res *res,
987                         struct rpc_task *task)
988 {
989         struct nfs4_session *session = nfs4_get_session(client);
990         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
991         struct nfs4_slot *slot;
992
993         /* slot already allocated? */
994         if (res->sr_slot != NULL)
995                 goto out_start;
996
997         if (session)
998                 tbl = &session->fc_slot_table;
999
1000         spin_lock(&tbl->slot_tbl_lock);
1001         /* The state manager will wait until the slot table is empty */
1002         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1003                 goto out_sleep;
1004
1005         slot = nfs4_alloc_slot(tbl);
1006         if (IS_ERR(slot)) {
1007                 if (slot == ERR_PTR(-ENOMEM))
1008                         goto out_sleep_timeout;
1009                 goto out_sleep;
1010         }
1011         spin_unlock(&tbl->slot_tbl_lock);
1012
1013         nfs4_sequence_attach_slot(args, res, slot);
1014
1015         trace_nfs4_setup_sequence(session, args);
1016 out_start:
1017         nfs41_sequence_res_init(res);
1018         rpc_call_start(task);
1019         return 0;
1020 out_sleep_timeout:
1021         /* Try again in 1/4 second */
1022         if (args->sa_privileged)
1023                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1024                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1025         else
1026                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1027                                 NULL, jiffies + (HZ >> 2));
1028         spin_unlock(&tbl->slot_tbl_lock);
1029         return -EAGAIN;
1030 out_sleep:
1031         if (args->sa_privileged)
1032                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1033                                 RPC_PRIORITY_PRIVILEGED);
1034         else
1035                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1036         spin_unlock(&tbl->slot_tbl_lock);
1037         return -EAGAIN;
1038 }
1039 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1040
1041 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1042 {
1043         struct nfs4_call_sync_data *data = calldata;
1044         nfs4_setup_sequence(data->seq_server->nfs_client,
1045                                 data->seq_args, data->seq_res, task);
1046 }
1047
1048 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1049 {
1050         struct nfs4_call_sync_data *data = calldata;
1051         nfs4_sequence_done(task, data->seq_res);
1052 }
1053
1054 static const struct rpc_call_ops nfs40_call_sync_ops = {
1055         .rpc_call_prepare = nfs40_call_sync_prepare,
1056         .rpc_call_done = nfs40_call_sync_done,
1057 };
1058
1059 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1060                                    struct nfs_server *server,
1061                                    struct rpc_message *msg,
1062                                    struct nfs4_sequence_args *args,
1063                                    struct nfs4_sequence_res *res)
1064 {
1065         int ret;
1066         struct rpc_task *task;
1067         struct nfs_client *clp = server->nfs_client;
1068         struct nfs4_call_sync_data data = {
1069                 .seq_server = server,
1070                 .seq_args = args,
1071                 .seq_res = res,
1072         };
1073         struct rpc_task_setup task_setup = {
1074                 .rpc_client = clnt,
1075                 .rpc_message = msg,
1076                 .callback_ops = clp->cl_mvops->call_sync_ops,
1077                 .callback_data = &data
1078         };
1079
1080         task = rpc_run_task(&task_setup);
1081         if (IS_ERR(task))
1082                 ret = PTR_ERR(task);
1083         else {
1084                 ret = task->tk_status;
1085                 rpc_put_task(task);
1086         }
1087         return ret;
1088 }
1089
1090 int nfs4_call_sync(struct rpc_clnt *clnt,
1091                    struct nfs_server *server,
1092                    struct rpc_message *msg,
1093                    struct nfs4_sequence_args *args,
1094                    struct nfs4_sequence_res *res,
1095                    int cache_reply)
1096 {
1097         nfs4_init_sequence(args, res, cache_reply, 0);
1098         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1099 }
1100
1101 static void
1102 nfs4_inc_nlink_locked(struct inode *inode)
1103 {
1104         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1105         inc_nlink(inode);
1106 }
1107
1108 static void
1109 nfs4_dec_nlink_locked(struct inode *inode)
1110 {
1111         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1112         drop_nlink(inode);
1113 }
1114
1115 static void
1116 update_changeattr_locked(struct inode *dir, struct nfs4_change_info *cinfo,
1117                 unsigned long timestamp, unsigned long cache_validity)
1118 {
1119         struct nfs_inode *nfsi = NFS_I(dir);
1120
1121         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1122                 | NFS_INO_INVALID_MTIME
1123                 | NFS_INO_INVALID_DATA
1124                 | cache_validity;
1125         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(dir)) {
1126                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1127                 nfsi->attrtimeo_timestamp = jiffies;
1128         } else {
1129                 nfs_force_lookup_revalidate(dir);
1130                 if (cinfo->before != inode_peek_iversion_raw(dir))
1131                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1132                                 NFS_INO_INVALID_ACL;
1133         }
1134         inode_set_iversion_raw(dir, cinfo->after);
1135         nfsi->read_cache_jiffies = timestamp;
1136         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1137         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1138         nfs_fscache_invalidate(dir);
1139 }
1140
1141 static void
1142 update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1143                 unsigned long timestamp, unsigned long cache_validity)
1144 {
1145         spin_lock(&dir->i_lock);
1146         update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1147         spin_unlock(&dir->i_lock);
1148 }
1149
1150 struct nfs4_open_createattrs {
1151         struct nfs4_label *label;
1152         struct iattr *sattr;
1153         const __u32 verf[2];
1154 };
1155
1156 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1157                 int err, struct nfs4_exception *exception)
1158 {
1159         if (err != -EINVAL)
1160                 return false;
1161         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1162                 return false;
1163         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1164         exception->retry = 1;
1165         return true;
1166 }
1167
1168 static u32
1169 nfs4_map_atomic_open_share(struct nfs_server *server,
1170                 fmode_t fmode, int openflags)
1171 {
1172         u32 res = 0;
1173
1174         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1175         case FMODE_READ:
1176                 res = NFS4_SHARE_ACCESS_READ;
1177                 break;
1178         case FMODE_WRITE:
1179                 res = NFS4_SHARE_ACCESS_WRITE;
1180                 break;
1181         case FMODE_READ|FMODE_WRITE:
1182                 res = NFS4_SHARE_ACCESS_BOTH;
1183         }
1184         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1185                 goto out;
1186         /* Want no delegation if we're using O_DIRECT */
1187         if (openflags & O_DIRECT)
1188                 res |= NFS4_SHARE_WANT_NO_DELEG;
1189 out:
1190         return res;
1191 }
1192
1193 static enum open_claim_type4
1194 nfs4_map_atomic_open_claim(struct nfs_server *server,
1195                 enum open_claim_type4 claim)
1196 {
1197         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1198                 return claim;
1199         switch (claim) {
1200         default:
1201                 return claim;
1202         case NFS4_OPEN_CLAIM_FH:
1203                 return NFS4_OPEN_CLAIM_NULL;
1204         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1205                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1206         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1207                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1208         }
1209 }
1210
1211 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1212 {
1213         p->o_res.f_attr = &p->f_attr;
1214         p->o_res.f_label = p->f_label;
1215         p->o_res.seqid = p->o_arg.seqid;
1216         p->c_res.seqid = p->c_arg.seqid;
1217         p->o_res.server = p->o_arg.server;
1218         p->o_res.access_request = p->o_arg.access;
1219         nfs_fattr_init(&p->f_attr);
1220         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1221 }
1222
1223 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1224                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1225                 const struct nfs4_open_createattrs *c,
1226                 enum open_claim_type4 claim,
1227                 gfp_t gfp_mask)
1228 {
1229         struct dentry *parent = dget_parent(dentry);
1230         struct inode *dir = d_inode(parent);
1231         struct nfs_server *server = NFS_SERVER(dir);
1232         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1233         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1234         struct nfs4_opendata *p;
1235
1236         p = kzalloc(sizeof(*p), gfp_mask);
1237         if (p == NULL)
1238                 goto err;
1239
1240         p->f_label = nfs4_label_alloc(server, gfp_mask);
1241         if (IS_ERR(p->f_label))
1242                 goto err_free_p;
1243
1244         p->a_label = nfs4_label_alloc(server, gfp_mask);
1245         if (IS_ERR(p->a_label))
1246                 goto err_free_f;
1247
1248         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1249         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1250         if (IS_ERR(p->o_arg.seqid))
1251                 goto err_free_label;
1252         nfs_sb_active(dentry->d_sb);
1253         p->dentry = dget(dentry);
1254         p->dir = parent;
1255         p->owner = sp;
1256         atomic_inc(&sp->so_count);
1257         p->o_arg.open_flags = flags;
1258         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1259         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1260         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1261                         fmode, flags);
1262         if (flags & O_CREAT) {
1263                 p->o_arg.umask = current_umask();
1264                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1265                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1266                         p->o_arg.u.attrs = &p->attrs;
1267                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1268
1269                         memcpy(p->o_arg.u.verifier.data, c->verf,
1270                                         sizeof(p->o_arg.u.verifier.data));
1271                 }
1272         }
1273         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1274          * will return permission denied for all bits until close */
1275         if (!(flags & O_EXCL)) {
1276                 /* ask server to check for all possible rights as results
1277                  * are cached */
1278                 switch (p->o_arg.claim) {
1279                 default:
1280                         break;
1281                 case NFS4_OPEN_CLAIM_NULL:
1282                 case NFS4_OPEN_CLAIM_FH:
1283                         p->o_arg.access = NFS4_ACCESS_READ |
1284                                 NFS4_ACCESS_MODIFY |
1285                                 NFS4_ACCESS_EXTEND |
1286                                 NFS4_ACCESS_EXECUTE;
1287                 }
1288         }
1289         p->o_arg.clientid = server->nfs_client->cl_clientid;
1290         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1291         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1292         p->o_arg.name = &dentry->d_name;
1293         p->o_arg.server = server;
1294         p->o_arg.bitmask = nfs4_bitmask(server, label);
1295         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1296         switch (p->o_arg.claim) {
1297         case NFS4_OPEN_CLAIM_NULL:
1298         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1299         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1300                 p->o_arg.fh = NFS_FH(dir);
1301                 break;
1302         case NFS4_OPEN_CLAIM_PREVIOUS:
1303         case NFS4_OPEN_CLAIM_FH:
1304         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1305         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1306                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1307         }
1308         p->c_arg.fh = &p->o_res.fh;
1309         p->c_arg.stateid = &p->o_res.stateid;
1310         p->c_arg.seqid = p->o_arg.seqid;
1311         nfs4_init_opendata_res(p);
1312         kref_init(&p->kref);
1313         return p;
1314
1315 err_free_label:
1316         nfs4_label_free(p->a_label);
1317 err_free_f:
1318         nfs4_label_free(p->f_label);
1319 err_free_p:
1320         kfree(p);
1321 err:
1322         dput(parent);
1323         return NULL;
1324 }
1325
1326 static void nfs4_opendata_free(struct kref *kref)
1327 {
1328         struct nfs4_opendata *p = container_of(kref,
1329                         struct nfs4_opendata, kref);
1330         struct super_block *sb = p->dentry->d_sb;
1331
1332         nfs4_lgopen_release(p->lgp);
1333         nfs_free_seqid(p->o_arg.seqid);
1334         nfs4_sequence_free_slot(&p->o_res.seq_res);
1335         if (p->state != NULL)
1336                 nfs4_put_open_state(p->state);
1337         nfs4_put_state_owner(p->owner);
1338
1339         nfs4_label_free(p->a_label);
1340         nfs4_label_free(p->f_label);
1341
1342         dput(p->dir);
1343         dput(p->dentry);
1344         nfs_sb_deactive(sb);
1345         nfs_fattr_free_names(&p->f_attr);
1346         kfree(p->f_attr.mdsthreshold);
1347         kfree(p);
1348 }
1349
1350 static void nfs4_opendata_put(struct nfs4_opendata *p)
1351 {
1352         if (p != NULL)
1353                 kref_put(&p->kref, nfs4_opendata_free);
1354 }
1355
1356 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1357                 fmode_t fmode)
1358 {
1359         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1360         case FMODE_READ|FMODE_WRITE:
1361                 return state->n_rdwr != 0;
1362         case FMODE_WRITE:
1363                 return state->n_wronly != 0;
1364         case FMODE_READ:
1365                 return state->n_rdonly != 0;
1366         }
1367         WARN_ON_ONCE(1);
1368         return false;
1369 }
1370
1371 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1372                 int open_mode, enum open_claim_type4 claim)
1373 {
1374         int ret = 0;
1375
1376         if (open_mode & (O_EXCL|O_TRUNC))
1377                 goto out;
1378         switch (claim) {
1379         case NFS4_OPEN_CLAIM_NULL:
1380         case NFS4_OPEN_CLAIM_FH:
1381                 goto out;
1382         default:
1383                 break;
1384         }
1385         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1386                 case FMODE_READ:
1387                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1388                                 && state->n_rdonly != 0;
1389                         break;
1390                 case FMODE_WRITE:
1391                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1392                                 && state->n_wronly != 0;
1393                         break;
1394                 case FMODE_READ|FMODE_WRITE:
1395                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1396                                 && state->n_rdwr != 0;
1397         }
1398 out:
1399         return ret;
1400 }
1401
1402 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1403                 enum open_claim_type4 claim)
1404 {
1405         if (delegation == NULL)
1406                 return 0;
1407         if ((delegation->type & fmode) != fmode)
1408                 return 0;
1409         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1410                 return 0;
1411         switch (claim) {
1412         case NFS4_OPEN_CLAIM_NULL:
1413         case NFS4_OPEN_CLAIM_FH:
1414                 break;
1415         case NFS4_OPEN_CLAIM_PREVIOUS:
1416                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1417                         break;
1418                 /* Fall through */
1419         default:
1420                 return 0;
1421         }
1422         nfs_mark_delegation_referenced(delegation);
1423         return 1;
1424 }
1425
1426 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1427 {
1428         switch (fmode) {
1429                 case FMODE_WRITE:
1430                         state->n_wronly++;
1431                         break;
1432                 case FMODE_READ:
1433                         state->n_rdonly++;
1434                         break;
1435                 case FMODE_READ|FMODE_WRITE:
1436                         state->n_rdwr++;
1437         }
1438         nfs4_state_set_mode_locked(state, state->state | fmode);
1439 }
1440
1441 #ifdef CONFIG_NFS_V4_1
1442 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1443 {
1444         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1445                 return true;
1446         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1447                 return true;
1448         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1449                 return true;
1450         return false;
1451 }
1452 #endif /* CONFIG_NFS_V4_1 */
1453
1454 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1455 {
1456         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1457                 wake_up_all(&state->waitq);
1458 }
1459
1460 static void nfs_state_log_out_of_order_open_stateid(struct nfs4_state *state,
1461                 const nfs4_stateid *stateid)
1462 {
1463         u32 state_seqid = be32_to_cpu(state->open_stateid.seqid);
1464         u32 stateid_seqid = be32_to_cpu(stateid->seqid);
1465
1466         if (stateid_seqid == state_seqid + 1U ||
1467             (stateid_seqid == 1U && state_seqid == 0xffffffffU))
1468                 nfs_state_log_update_open_stateid(state);
1469         else
1470                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1471 }
1472
1473 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1474 {
1475         struct nfs_client *clp = state->owner->so_server->nfs_client;
1476         bool need_recover = false;
1477
1478         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1479                 need_recover = true;
1480         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1481                 need_recover = true;
1482         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1483                 need_recover = true;
1484         if (need_recover)
1485                 nfs4_state_mark_reclaim_nograce(clp, state);
1486 }
1487
1488 /*
1489  * Check for whether or not the caller may update the open stateid
1490  * to the value passed in by stateid.
1491  *
1492  * Note: This function relies heavily on the server implementing
1493  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1494  * correctly.
1495  * i.e. The stateid seqids have to be initialised to 1, and
1496  * are then incremented on every state transition.
1497  */
1498 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1499                 const nfs4_stateid *stateid)
1500 {
1501         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0 ||
1502             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1503                 if (stateid->seqid == cpu_to_be32(1))
1504                         nfs_state_log_update_open_stateid(state);
1505                 else
1506                         set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1507                 return true;
1508         }
1509
1510         if (nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1511                 nfs_state_log_out_of_order_open_stateid(state, stateid);
1512                 return true;
1513         }
1514         return false;
1515 }
1516
1517 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1518 {
1519         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1520                 return;
1521         if (state->n_wronly)
1522                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1523         if (state->n_rdonly)
1524                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1525         if (state->n_rdwr)
1526                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1527         set_bit(NFS_OPEN_STATE, &state->flags);
1528 }
1529
1530 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1531                 nfs4_stateid *stateid, fmode_t fmode)
1532 {
1533         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1534         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1535         case FMODE_WRITE:
1536                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1537                 break;
1538         case FMODE_READ:
1539                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1540                 break;
1541         case 0:
1542                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1543                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1544                 clear_bit(NFS_OPEN_STATE, &state->flags);
1545         }
1546         if (stateid == NULL)
1547                 return;
1548         /* Handle OPEN+OPEN_DOWNGRADE races */
1549         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1550             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1551                 nfs_resync_open_stateid_locked(state);
1552                 goto out;
1553         }
1554         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1555                 nfs4_stateid_copy(&state->stateid, stateid);
1556         nfs4_stateid_copy(&state->open_stateid, stateid);
1557         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1558 out:
1559         nfs_state_log_update_open_stateid(state);
1560 }
1561
1562 static void nfs_clear_open_stateid(struct nfs4_state *state,
1563         nfs4_stateid *arg_stateid,
1564         nfs4_stateid *stateid, fmode_t fmode)
1565 {
1566         write_seqlock(&state->seqlock);
1567         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1568         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1569                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1570         write_sequnlock(&state->seqlock);
1571         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1572                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1573 }
1574
1575 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1576                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1577         __must_hold(&state->owner->so_lock)
1578         __must_hold(&state->seqlock)
1579         __must_hold(RCU)
1580
1581 {
1582         DEFINE_WAIT(wait);
1583         int status = 0;
1584         for (;;) {
1585
1586                 if (!nfs_need_update_open_stateid(state, stateid))
1587                         return;
1588                 if (!test_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1589                         break;
1590                 if (status)
1591                         break;
1592                 /* Rely on seqids for serialisation with NFSv4.0 */
1593                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1594                         break;
1595
1596                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1597                 /*
1598                  * Ensure we process the state changes in the same order
1599                  * in which the server processed them by delaying the
1600                  * update of the stateid until we are in sequence.
1601                  */
1602                 write_sequnlock(&state->seqlock);
1603                 spin_unlock(&state->owner->so_lock);
1604                 rcu_read_unlock();
1605                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1606                 if (!signal_pending(current)) {
1607                         if (schedule_timeout(5*HZ) == 0)
1608                                 status = -EAGAIN;
1609                         else
1610                                 status = 0;
1611                 } else
1612                         status = -EINTR;
1613                 finish_wait(&state->waitq, &wait);
1614                 rcu_read_lock();
1615                 spin_lock(&state->owner->so_lock);
1616                 write_seqlock(&state->seqlock);
1617         }
1618
1619         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1620             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1621                 nfs4_stateid_copy(freeme, &state->open_stateid);
1622                 nfs_test_and_clear_all_open_stateid(state);
1623         }
1624
1625         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1626                 nfs4_stateid_copy(&state->stateid, stateid);
1627         nfs4_stateid_copy(&state->open_stateid, stateid);
1628         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1629         nfs_state_log_update_open_stateid(state);
1630 }
1631
1632 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1633                 const nfs4_stateid *open_stateid,
1634                 fmode_t fmode,
1635                 nfs4_stateid *freeme)
1636 {
1637         /*
1638          * Protect the call to nfs4_state_set_mode_locked and
1639          * serialise the stateid update
1640          */
1641         write_seqlock(&state->seqlock);
1642         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1643         switch (fmode) {
1644         case FMODE_READ:
1645                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1646                 break;
1647         case FMODE_WRITE:
1648                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1649                 break;
1650         case FMODE_READ|FMODE_WRITE:
1651                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1652         }
1653         set_bit(NFS_OPEN_STATE, &state->flags);
1654         write_sequnlock(&state->seqlock);
1655 }
1656
1657 static void nfs_state_set_delegation(struct nfs4_state *state,
1658                 const nfs4_stateid *deleg_stateid,
1659                 fmode_t fmode)
1660 {
1661         /*
1662          * Protect the call to nfs4_state_set_mode_locked and
1663          * serialise the stateid update
1664          */
1665         write_seqlock(&state->seqlock);
1666         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1667         set_bit(NFS_DELEGATED_STATE, &state->flags);
1668         write_sequnlock(&state->seqlock);
1669 }
1670
1671 static void nfs_state_clear_delegation(struct nfs4_state *state)
1672 {
1673         write_seqlock(&state->seqlock);
1674         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1675         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1676         write_sequnlock(&state->seqlock);
1677 }
1678
1679 static int update_open_stateid(struct nfs4_state *state,
1680                 const nfs4_stateid *open_stateid,
1681                 const nfs4_stateid *delegation,
1682                 fmode_t fmode)
1683 {
1684         struct nfs_server *server = NFS_SERVER(state->inode);
1685         struct nfs_client *clp = server->nfs_client;
1686         struct nfs_inode *nfsi = NFS_I(state->inode);
1687         struct nfs_delegation *deleg_cur;
1688         nfs4_stateid freeme = { };
1689         int ret = 0;
1690
1691         fmode &= (FMODE_READ|FMODE_WRITE);
1692
1693         rcu_read_lock();
1694         spin_lock(&state->owner->so_lock);
1695         if (open_stateid != NULL) {
1696                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1697                 ret = 1;
1698         }
1699
1700         deleg_cur = rcu_dereference(nfsi->delegation);
1701         if (deleg_cur == NULL)
1702                 goto no_delegation;
1703
1704         spin_lock(&deleg_cur->lock);
1705         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1706            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1707             (deleg_cur->type & fmode) != fmode)
1708                 goto no_delegation_unlock;
1709
1710         if (delegation == NULL)
1711                 delegation = &deleg_cur->stateid;
1712         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1713                 goto no_delegation_unlock;
1714
1715         nfs_mark_delegation_referenced(deleg_cur);
1716         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1717         ret = 1;
1718 no_delegation_unlock:
1719         spin_unlock(&deleg_cur->lock);
1720 no_delegation:
1721         if (ret)
1722                 update_open_stateflags(state, fmode);
1723         spin_unlock(&state->owner->so_lock);
1724         rcu_read_unlock();
1725
1726         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1727                 nfs4_schedule_state_manager(clp);
1728         if (freeme.type != 0)
1729                 nfs4_test_and_free_stateid(server, &freeme,
1730                                 state->owner->so_cred);
1731
1732         return ret;
1733 }
1734
1735 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1736                 const nfs4_stateid *stateid)
1737 {
1738         struct nfs4_state *state = lsp->ls_state;
1739         bool ret = false;
1740
1741         spin_lock(&state->state_lock);
1742         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1743                 goto out_noupdate;
1744         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1745                 goto out_noupdate;
1746         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1747         ret = true;
1748 out_noupdate:
1749         spin_unlock(&state->state_lock);
1750         return ret;
1751 }
1752
1753 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1754 {
1755         struct nfs_delegation *delegation;
1756
1757         fmode &= FMODE_READ|FMODE_WRITE;
1758         rcu_read_lock();
1759         delegation = rcu_dereference(NFS_I(inode)->delegation);
1760         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1761                 rcu_read_unlock();
1762                 return;
1763         }
1764         rcu_read_unlock();
1765         nfs4_inode_return_delegation(inode);
1766 }
1767
1768 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1769 {
1770         struct nfs4_state *state = opendata->state;
1771         struct nfs_inode *nfsi = NFS_I(state->inode);
1772         struct nfs_delegation *delegation;
1773         int open_mode = opendata->o_arg.open_flags;
1774         fmode_t fmode = opendata->o_arg.fmode;
1775         enum open_claim_type4 claim = opendata->o_arg.claim;
1776         nfs4_stateid stateid;
1777         int ret = -EAGAIN;
1778
1779         for (;;) {
1780                 spin_lock(&state->owner->so_lock);
1781                 if (can_open_cached(state, fmode, open_mode, claim)) {
1782                         update_open_stateflags(state, fmode);
1783                         spin_unlock(&state->owner->so_lock);
1784                         goto out_return_state;
1785                 }
1786                 spin_unlock(&state->owner->so_lock);
1787                 rcu_read_lock();
1788                 delegation = rcu_dereference(nfsi->delegation);
1789                 if (!can_open_delegated(delegation, fmode, claim)) {
1790                         rcu_read_unlock();
1791                         break;
1792                 }
1793                 /* Save the delegation */
1794                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1795                 rcu_read_unlock();
1796                 nfs_release_seqid(opendata->o_arg.seqid);
1797                 if (!opendata->is_recover) {
1798                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1799                         if (ret != 0)
1800                                 goto out;
1801                 }
1802                 ret = -EAGAIN;
1803
1804                 /* Try to update the stateid using the delegation */
1805                 if (update_open_stateid(state, NULL, &stateid, fmode))
1806                         goto out_return_state;
1807         }
1808 out:
1809         return ERR_PTR(ret);
1810 out_return_state:
1811         refcount_inc(&state->count);
1812         return state;
1813 }
1814
1815 static void
1816 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1817 {
1818         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1819         struct nfs_delegation *delegation;
1820         int delegation_flags = 0;
1821
1822         rcu_read_lock();
1823         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1824         if (delegation)
1825                 delegation_flags = delegation->flags;
1826         rcu_read_unlock();
1827         switch (data->o_arg.claim) {
1828         default:
1829                 break;
1830         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1831         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1832                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1833                                    "returning a delegation for "
1834                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1835                                    clp->cl_hostname);
1836                 return;
1837         }
1838         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1839                 nfs_inode_set_delegation(state->inode,
1840                                 data->owner->so_cred,
1841                                 data->o_res.delegation_type,
1842                                 &data->o_res.delegation,
1843                                 data->o_res.pagemod_limit);
1844         else
1845                 nfs_inode_reclaim_delegation(state->inode,
1846                                 data->owner->so_cred,
1847                                 data->o_res.delegation_type,
1848                                 &data->o_res.delegation,
1849                                 data->o_res.pagemod_limit);
1850
1851         if (data->o_res.do_recall)
1852                 nfs_async_inode_return_delegation(state->inode,
1853                                                   &data->o_res.delegation);
1854 }
1855
1856 /*
1857  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1858  * and update the nfs4_state.
1859  */
1860 static struct nfs4_state *
1861 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1862 {
1863         struct inode *inode = data->state->inode;
1864         struct nfs4_state *state = data->state;
1865         int ret;
1866
1867         if (!data->rpc_done) {
1868                 if (data->rpc_status)
1869                         return ERR_PTR(data->rpc_status);
1870                 /* cached opens have already been processed */
1871                 goto update;
1872         }
1873
1874         ret = nfs_refresh_inode(inode, &data->f_attr);
1875         if (ret)
1876                 return ERR_PTR(ret);
1877
1878         if (data->o_res.delegation_type != 0)
1879                 nfs4_opendata_check_deleg(data, state);
1880 update:
1881         update_open_stateid(state, &data->o_res.stateid, NULL,
1882                             data->o_arg.fmode);
1883         refcount_inc(&state->count);
1884
1885         return state;
1886 }
1887
1888 static struct inode *
1889 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1890 {
1891         struct inode *inode;
1892
1893         switch (data->o_arg.claim) {
1894         case NFS4_OPEN_CLAIM_NULL:
1895         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1896         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1897                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1898                         return ERR_PTR(-EAGAIN);
1899                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1900                                 &data->f_attr, data->f_label);
1901                 break;
1902         default:
1903                 inode = d_inode(data->dentry);
1904                 ihold(inode);
1905                 nfs_refresh_inode(inode, &data->f_attr);
1906         }
1907         return inode;
1908 }
1909
1910 static struct nfs4_state *
1911 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
1912 {
1913         struct nfs4_state *state;
1914         struct inode *inode;
1915
1916         inode = nfs4_opendata_get_inode(data);
1917         if (IS_ERR(inode))
1918                 return ERR_CAST(inode);
1919         if (data->state != NULL && data->state->inode == inode) {
1920                 state = data->state;
1921                 refcount_inc(&state->count);
1922         } else
1923                 state = nfs4_get_open_state(inode, data->owner);
1924         iput(inode);
1925         if (state == NULL)
1926                 state = ERR_PTR(-ENOMEM);
1927         return state;
1928 }
1929
1930 static struct nfs4_state *
1931 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1932 {
1933         struct nfs4_state *state;
1934
1935         if (!data->rpc_done) {
1936                 state = nfs4_try_open_cached(data);
1937                 trace_nfs4_cached_open(data->state);
1938                 goto out;
1939         }
1940
1941         state = nfs4_opendata_find_nfs4_state(data);
1942         if (IS_ERR(state))
1943                 goto out;
1944
1945         if (data->o_res.delegation_type != 0)
1946                 nfs4_opendata_check_deleg(data, state);
1947         update_open_stateid(state, &data->o_res.stateid, NULL,
1948                         data->o_arg.fmode);
1949 out:
1950         nfs_release_seqid(data->o_arg.seqid);
1951         return state;
1952 }
1953
1954 static struct nfs4_state *
1955 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1956 {
1957         struct nfs4_state *ret;
1958
1959         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1960                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1961         else
1962                 ret = _nfs4_opendata_to_nfs4_state(data);
1963         nfs4_sequence_free_slot(&data->o_res.seq_res);
1964         return ret;
1965 }
1966
1967 static struct nfs_open_context *
1968 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
1969 {
1970         struct nfs_inode *nfsi = NFS_I(state->inode);
1971         struct nfs_open_context *ctx;
1972
1973         rcu_read_lock();
1974         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
1975                 if (ctx->state != state)
1976                         continue;
1977                 if ((ctx->mode & mode) != mode)
1978                         continue;
1979                 if (!get_nfs_open_context(ctx))
1980                         continue;
1981                 rcu_read_unlock();
1982                 return ctx;
1983         }
1984         rcu_read_unlock();
1985         return ERR_PTR(-ENOENT);
1986 }
1987
1988 static struct nfs_open_context *
1989 nfs4_state_find_open_context(struct nfs4_state *state)
1990 {
1991         struct nfs_open_context *ctx;
1992
1993         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
1994         if (!IS_ERR(ctx))
1995                 return ctx;
1996         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
1997         if (!IS_ERR(ctx))
1998                 return ctx;
1999         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2000 }
2001
2002 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2003                 struct nfs4_state *state, enum open_claim_type4 claim)
2004 {
2005         struct nfs4_opendata *opendata;
2006
2007         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2008                         NULL, claim, GFP_NOFS);
2009         if (opendata == NULL)
2010                 return ERR_PTR(-ENOMEM);
2011         opendata->state = state;
2012         refcount_inc(&state->count);
2013         return opendata;
2014 }
2015
2016 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2017                 fmode_t fmode)
2018 {
2019         struct nfs4_state *newstate;
2020         int ret;
2021
2022         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2023                 return 0;
2024         opendata->o_arg.open_flags = 0;
2025         opendata->o_arg.fmode = fmode;
2026         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2027                         NFS_SB(opendata->dentry->d_sb),
2028                         fmode, 0);
2029         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2030         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2031         nfs4_init_opendata_res(opendata);
2032         ret = _nfs4_recover_proc_open(opendata);
2033         if (ret != 0)
2034                 return ret; 
2035         newstate = nfs4_opendata_to_nfs4_state(opendata);
2036         if (IS_ERR(newstate))
2037                 return PTR_ERR(newstate);
2038         if (newstate != opendata->state)
2039                 ret = -ESTALE;
2040         nfs4_close_state(newstate, fmode);
2041         return ret;
2042 }
2043
2044 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2045 {
2046         int ret;
2047
2048         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
2049         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2050         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2051         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2052         /* memory barrier prior to reading state->n_* */
2053         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2054         clear_bit(NFS_OPEN_STATE, &state->flags);
2055         smp_rmb();
2056         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2057         if (ret != 0)
2058                 return ret;
2059         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2060         if (ret != 0)
2061                 return ret;
2062         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2063         if (ret != 0)
2064                 return ret;
2065         /*
2066          * We may have performed cached opens for all three recoveries.
2067          * Check if we need to update the current stateid.
2068          */
2069         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2070             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2071                 write_seqlock(&state->seqlock);
2072                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2073                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2074                 write_sequnlock(&state->seqlock);
2075         }
2076         return 0;
2077 }
2078
2079 /*
2080  * OPEN_RECLAIM:
2081  *      reclaim state on the server after a reboot.
2082  */
2083 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2084 {
2085         struct nfs_delegation *delegation;
2086         struct nfs4_opendata *opendata;
2087         fmode_t delegation_type = 0;
2088         int status;
2089
2090         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2091                         NFS4_OPEN_CLAIM_PREVIOUS);
2092         if (IS_ERR(opendata))
2093                 return PTR_ERR(opendata);
2094         rcu_read_lock();
2095         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2096         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2097                 delegation_type = delegation->type;
2098         rcu_read_unlock();
2099         opendata->o_arg.u.delegation_type = delegation_type;
2100         status = nfs4_open_recover(opendata, state);
2101         nfs4_opendata_put(opendata);
2102         return status;
2103 }
2104
2105 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2106 {
2107         struct nfs_server *server = NFS_SERVER(state->inode);
2108         struct nfs4_exception exception = { };
2109         int err;
2110         do {
2111                 err = _nfs4_do_open_reclaim(ctx, state);
2112                 trace_nfs4_open_reclaim(ctx, 0, err);
2113                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2114                         continue;
2115                 if (err != -NFS4ERR_DELAY)
2116                         break;
2117                 nfs4_handle_exception(server, err, &exception);
2118         } while (exception.retry);
2119         return err;
2120 }
2121
2122 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2123 {
2124         struct nfs_open_context *ctx;
2125         int ret;
2126
2127         ctx = nfs4_state_find_open_context(state);
2128         if (IS_ERR(ctx))
2129                 return -EAGAIN;
2130         ret = nfs4_do_open_reclaim(ctx, state);
2131         put_nfs_open_context(ctx);
2132         return ret;
2133 }
2134
2135 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, struct file_lock *fl, int err)
2136 {
2137         switch (err) {
2138                 default:
2139                         printk(KERN_ERR "NFS: %s: unhandled error "
2140                                         "%d.\n", __func__, err);
2141                 case 0:
2142                 case -ENOENT:
2143                 case -EAGAIN:
2144                 case -ESTALE:
2145                         break;
2146                 case -NFS4ERR_BADSESSION:
2147                 case -NFS4ERR_BADSLOT:
2148                 case -NFS4ERR_BAD_HIGH_SLOT:
2149                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2150                 case -NFS4ERR_DEADSESSION:
2151                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2152                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
2153                         return -EAGAIN;
2154                 case -NFS4ERR_STALE_CLIENTID:
2155                 case -NFS4ERR_STALE_STATEID:
2156                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2157                         /* Don't recall a delegation if it was lost */
2158                         nfs4_schedule_lease_recovery(server->nfs_client);
2159                         return -EAGAIN;
2160                 case -NFS4ERR_MOVED:
2161                         nfs4_schedule_migration_recovery(server);
2162                         return -EAGAIN;
2163                 case -NFS4ERR_LEASE_MOVED:
2164                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2165                         return -EAGAIN;
2166                 case -NFS4ERR_DELEG_REVOKED:
2167                 case -NFS4ERR_ADMIN_REVOKED:
2168                 case -NFS4ERR_EXPIRED:
2169                 case -NFS4ERR_BAD_STATEID:
2170                 case -NFS4ERR_OPENMODE:
2171                         nfs_inode_find_state_and_recover(state->inode,
2172                                         stateid);
2173                         nfs4_schedule_stateid_recovery(server, state);
2174                         return -EAGAIN;
2175                 case -NFS4ERR_DELAY:
2176                 case -NFS4ERR_GRACE:
2177                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2178                         ssleep(1);
2179                         return -EAGAIN;
2180                 case -ENOMEM:
2181                 case -NFS4ERR_DENIED:
2182                         if (fl) {
2183                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2184                                 if (lsp)
2185                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2186                         }
2187                         return 0;
2188         }
2189         return err;
2190 }
2191
2192 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2193                 struct nfs4_state *state, const nfs4_stateid *stateid,
2194                 fmode_t type)
2195 {
2196         struct nfs_server *server = NFS_SERVER(state->inode);
2197         struct nfs4_opendata *opendata;
2198         int err = 0;
2199
2200         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2201                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2202         if (IS_ERR(opendata))
2203                 return PTR_ERR(opendata);
2204         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2205         nfs_state_clear_delegation(state);
2206         switch (type & (FMODE_READ|FMODE_WRITE)) {
2207         case FMODE_READ|FMODE_WRITE:
2208         case FMODE_WRITE:
2209                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2210                 if (err)
2211                         break;
2212                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2213                 if (err)
2214                         break;
2215                 /* Fall through */
2216         case FMODE_READ:
2217                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2218         }
2219         nfs4_opendata_put(opendata);
2220         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2221 }
2222
2223 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2224 {
2225         struct nfs4_opendata *data = calldata;
2226
2227         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2228                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2229 }
2230
2231 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2232 {
2233         struct nfs4_opendata *data = calldata;
2234
2235         nfs40_sequence_done(task, &data->c_res.seq_res);
2236
2237         data->rpc_status = task->tk_status;
2238         if (data->rpc_status == 0) {
2239                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2240                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2241                 renew_lease(data->o_res.server, data->timestamp);
2242                 data->rpc_done = true;
2243         }
2244 }
2245
2246 static void nfs4_open_confirm_release(void *calldata)
2247 {
2248         struct nfs4_opendata *data = calldata;
2249         struct nfs4_state *state = NULL;
2250
2251         /* If this request hasn't been cancelled, do nothing */
2252         if (!data->cancelled)
2253                 goto out_free;
2254         /* In case of error, no cleanup! */
2255         if (!data->rpc_done)
2256                 goto out_free;
2257         state = nfs4_opendata_to_nfs4_state(data);
2258         if (!IS_ERR(state))
2259                 nfs4_close_state(state, data->o_arg.fmode);
2260 out_free:
2261         nfs4_opendata_put(data);
2262 }
2263
2264 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2265         .rpc_call_prepare = nfs4_open_confirm_prepare,
2266         .rpc_call_done = nfs4_open_confirm_done,
2267         .rpc_release = nfs4_open_confirm_release,
2268 };
2269
2270 /*
2271  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2272  */
2273 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2274 {
2275         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2276         struct rpc_task *task;
2277         struct  rpc_message msg = {
2278                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2279                 .rpc_argp = &data->c_arg,
2280                 .rpc_resp = &data->c_res,
2281                 .rpc_cred = data->owner->so_cred,
2282         };
2283         struct rpc_task_setup task_setup_data = {
2284                 .rpc_client = server->client,
2285                 .rpc_message = &msg,
2286                 .callback_ops = &nfs4_open_confirm_ops,
2287                 .callback_data = data,
2288                 .workqueue = nfsiod_workqueue,
2289                 .flags = RPC_TASK_ASYNC,
2290         };
2291         int status;
2292
2293         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2294                                 data->is_recover);
2295         kref_get(&data->kref);
2296         data->rpc_done = false;
2297         data->rpc_status = 0;
2298         data->timestamp = jiffies;
2299         task = rpc_run_task(&task_setup_data);
2300         if (IS_ERR(task))
2301                 return PTR_ERR(task);
2302         status = rpc_wait_for_completion_task(task);
2303         if (status != 0) {
2304                 data->cancelled = true;
2305                 smp_wmb();
2306         } else
2307                 status = data->rpc_status;
2308         rpc_put_task(task);
2309         return status;
2310 }
2311
2312 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2313 {
2314         struct nfs4_opendata *data = calldata;
2315         struct nfs4_state_owner *sp = data->owner;
2316         struct nfs_client *clp = sp->so_server->nfs_client;
2317         enum open_claim_type4 claim = data->o_arg.claim;
2318
2319         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2320                 goto out_wait;
2321         /*
2322          * Check if we still need to send an OPEN call, or if we can use
2323          * a delegation instead.
2324          */
2325         if (data->state != NULL) {
2326                 struct nfs_delegation *delegation;
2327
2328                 if (can_open_cached(data->state, data->o_arg.fmode,
2329                                         data->o_arg.open_flags, claim))
2330                         goto out_no_action;
2331                 rcu_read_lock();
2332                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2333                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2334                         goto unlock_no_action;
2335                 rcu_read_unlock();
2336         }
2337         /* Update client id. */
2338         data->o_arg.clientid = clp->cl_clientid;
2339         switch (claim) {
2340         default:
2341                 break;
2342         case NFS4_OPEN_CLAIM_PREVIOUS:
2343         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2344         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2345                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2346                 /* Fall through */
2347         case NFS4_OPEN_CLAIM_FH:
2348                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2349         }
2350         data->timestamp = jiffies;
2351         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2352                                 &data->o_arg.seq_args,
2353                                 &data->o_res.seq_res,
2354                                 task) != 0)
2355                 nfs_release_seqid(data->o_arg.seqid);
2356
2357         /* Set the create mode (note dependency on the session type) */
2358         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2359         if (data->o_arg.open_flags & O_EXCL) {
2360                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2361                 if (nfs4_has_persistent_session(clp))
2362                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2363                 else if (clp->cl_mvops->minor_version > 0)
2364                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2365         }
2366         return;
2367 unlock_no_action:
2368         trace_nfs4_cached_open(data->state);
2369         rcu_read_unlock();
2370 out_no_action:
2371         task->tk_action = NULL;
2372 out_wait:
2373         nfs4_sequence_done(task, &data->o_res.seq_res);
2374 }
2375
2376 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2377 {
2378         struct nfs4_opendata *data = calldata;
2379
2380         data->rpc_status = task->tk_status;
2381
2382         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2383                 return;
2384
2385         if (task->tk_status == 0) {
2386                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2387                         switch (data->o_res.f_attr->mode & S_IFMT) {
2388                         case S_IFREG:
2389                                 break;
2390                         case S_IFLNK:
2391                                 data->rpc_status = -ELOOP;
2392                                 break;
2393                         case S_IFDIR:
2394                                 data->rpc_status = -EISDIR;
2395                                 break;
2396                         default:
2397                                 data->rpc_status = -ENOTDIR;
2398                         }
2399                 }
2400                 renew_lease(data->o_res.server, data->timestamp);
2401                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2402                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2403         }
2404         data->rpc_done = true;
2405 }
2406
2407 static void nfs4_open_release(void *calldata)
2408 {
2409         struct nfs4_opendata *data = calldata;
2410         struct nfs4_state *state = NULL;
2411
2412         /* If this request hasn't been cancelled, do nothing */
2413         if (!data->cancelled)
2414                 goto out_free;
2415         /* In case of error, no cleanup! */
2416         if (data->rpc_status != 0 || !data->rpc_done)
2417                 goto out_free;
2418         /* In case we need an open_confirm, no cleanup! */
2419         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2420                 goto out_free;
2421         state = nfs4_opendata_to_nfs4_state(data);
2422         if (!IS_ERR(state))
2423                 nfs4_close_state(state, data->o_arg.fmode);
2424 out_free:
2425         nfs4_opendata_put(data);
2426 }
2427
2428 static const struct rpc_call_ops nfs4_open_ops = {
2429         .rpc_call_prepare = nfs4_open_prepare,
2430         .rpc_call_done = nfs4_open_done,
2431         .rpc_release = nfs4_open_release,
2432 };
2433
2434 static int nfs4_run_open_task(struct nfs4_opendata *data,
2435                               struct nfs_open_context *ctx)
2436 {
2437         struct inode *dir = d_inode(data->dir);
2438         struct nfs_server *server = NFS_SERVER(dir);
2439         struct nfs_openargs *o_arg = &data->o_arg;
2440         struct nfs_openres *o_res = &data->o_res;
2441         struct rpc_task *task;
2442         struct rpc_message msg = {
2443                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2444                 .rpc_argp = o_arg,
2445                 .rpc_resp = o_res,
2446                 .rpc_cred = data->owner->so_cred,
2447         };
2448         struct rpc_task_setup task_setup_data = {
2449                 .rpc_client = server->client,
2450                 .rpc_message = &msg,
2451                 .callback_ops = &nfs4_open_ops,
2452                 .callback_data = data,
2453                 .workqueue = nfsiod_workqueue,
2454                 .flags = RPC_TASK_ASYNC,
2455         };
2456         int status;
2457
2458         kref_get(&data->kref);
2459         data->rpc_done = false;
2460         data->rpc_status = 0;
2461         data->cancelled = false;
2462         data->is_recover = false;
2463         if (!ctx) {
2464                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2465                 data->is_recover = true;
2466         } else {
2467                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2468                 pnfs_lgopen_prepare(data, ctx);
2469         }
2470         task = rpc_run_task(&task_setup_data);
2471         if (IS_ERR(task))
2472                 return PTR_ERR(task);
2473         status = rpc_wait_for_completion_task(task);
2474         if (status != 0) {
2475                 data->cancelled = true;
2476                 smp_wmb();
2477         } else
2478                 status = data->rpc_status;
2479         rpc_put_task(task);
2480
2481         return status;
2482 }
2483
2484 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2485 {
2486         struct inode *dir = d_inode(data->dir);
2487         struct nfs_openres *o_res = &data->o_res;
2488         int status;
2489
2490         status = nfs4_run_open_task(data, NULL);
2491         if (status != 0 || !data->rpc_done)
2492                 return status;
2493
2494         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2495
2496         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2497                 status = _nfs4_proc_open_confirm(data);
2498
2499         return status;
2500 }
2501
2502 /*
2503  * Additional permission checks in order to distinguish between an
2504  * open for read, and an open for execute. This works around the
2505  * fact that NFSv4 OPEN treats read and execute permissions as being
2506  * the same.
2507  * Note that in the non-execute case, we want to turn off permission
2508  * checking if we just created a new file (POSIX open() semantics).
2509  */
2510 static int nfs4_opendata_access(const struct cred *cred,
2511                                 struct nfs4_opendata *opendata,
2512                                 struct nfs4_state *state, fmode_t fmode,
2513                                 int openflags)
2514 {
2515         struct nfs_access_entry cache;
2516         u32 mask, flags;
2517
2518         /* access call failed or for some reason the server doesn't
2519          * support any access modes -- defer access call until later */
2520         if (opendata->o_res.access_supported == 0)
2521                 return 0;
2522
2523         mask = 0;
2524         /*
2525          * Use openflags to check for exec, because fmode won't
2526          * always have FMODE_EXEC set when file open for exec.
2527          */
2528         if (openflags & __FMODE_EXEC) {
2529                 /* ONLY check for exec rights */
2530                 if (S_ISDIR(state->inode->i_mode))
2531                         mask = NFS4_ACCESS_LOOKUP;
2532                 else
2533                         mask = NFS4_ACCESS_EXECUTE;
2534         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2535                 mask = NFS4_ACCESS_READ;
2536
2537         cache.cred = cred;
2538         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2539         nfs_access_add_cache(state->inode, &cache);
2540
2541         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2542         if ((mask & ~cache.mask & flags) == 0)
2543                 return 0;
2544
2545         return -EACCES;
2546 }
2547
2548 /*
2549  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2550  */
2551 static int _nfs4_proc_open(struct nfs4_opendata *data,
2552                            struct nfs_open_context *ctx)
2553 {
2554         struct inode *dir = d_inode(data->dir);
2555         struct nfs_server *server = NFS_SERVER(dir);
2556         struct nfs_openargs *o_arg = &data->o_arg;
2557         struct nfs_openres *o_res = &data->o_res;
2558         int status;
2559
2560         status = nfs4_run_open_task(data, ctx);
2561         if (!data->rpc_done)
2562                 return status;
2563         if (status != 0) {
2564                 if (status == -NFS4ERR_BADNAME &&
2565                                 !(o_arg->open_flags & O_CREAT))
2566                         return -ENOENT;
2567                 return status;
2568         }
2569
2570         nfs_fattr_map_and_free_names(server, &data->f_attr);
2571
2572         if (o_arg->open_flags & O_CREAT) {
2573                 if (o_arg->open_flags & O_EXCL)
2574                         data->file_created = true;
2575                 else if (o_res->cinfo.before != o_res->cinfo.after)
2576                         data->file_created = true;
2577                 if (data->file_created ||
2578                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2579                         update_changeattr(dir, &o_res->cinfo,
2580                                         o_res->f_attr->time_start, 0);
2581         }
2582         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2583                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2584         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2585                 status = _nfs4_proc_open_confirm(data);
2586                 if (status != 0)
2587                         return status;
2588         }
2589         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2590                 nfs4_sequence_free_slot(&o_res->seq_res);
2591                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2592                                 o_res->f_label, NULL);
2593         }
2594         return 0;
2595 }
2596
2597 /*
2598  * OPEN_EXPIRED:
2599  *      reclaim state on the server after a network partition.
2600  *      Assumes caller holds the appropriate lock
2601  */
2602 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2603 {
2604         struct nfs4_opendata *opendata;
2605         int ret;
2606
2607         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2608                         NFS4_OPEN_CLAIM_FH);
2609         if (IS_ERR(opendata))
2610                 return PTR_ERR(opendata);
2611         ret = nfs4_open_recover(opendata, state);
2612         if (ret == -ESTALE)
2613                 d_drop(ctx->dentry);
2614         nfs4_opendata_put(opendata);
2615         return ret;
2616 }
2617
2618 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2619 {
2620         struct nfs_server *server = NFS_SERVER(state->inode);
2621         struct nfs4_exception exception = { };
2622         int err;
2623
2624         do {
2625                 err = _nfs4_open_expired(ctx, state);
2626                 trace_nfs4_open_expired(ctx, 0, err);
2627                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2628                         continue;
2629                 switch (err) {
2630                 default:
2631                         goto out;
2632                 case -NFS4ERR_GRACE:
2633                 case -NFS4ERR_DELAY:
2634                         nfs4_handle_exception(server, err, &exception);
2635                         err = 0;
2636                 }
2637         } while (exception.retry);
2638 out:
2639         return err;
2640 }
2641
2642 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2643 {
2644         struct nfs_open_context *ctx;
2645         int ret;
2646
2647         ctx = nfs4_state_find_open_context(state);
2648         if (IS_ERR(ctx))
2649                 return -EAGAIN;
2650         ret = nfs4_do_open_expired(ctx, state);
2651         put_nfs_open_context(ctx);
2652         return ret;
2653 }
2654
2655 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2656                 const nfs4_stateid *stateid)
2657 {
2658         nfs_remove_bad_delegation(state->inode, stateid);
2659         nfs_state_clear_delegation(state);
2660 }
2661
2662 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2663 {
2664         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2665                 nfs_finish_clear_delegation_stateid(state, NULL);
2666 }
2667
2668 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2669 {
2670         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2671         nfs40_clear_delegation_stateid(state);
2672         return nfs4_open_expired(sp, state);
2673 }
2674
2675 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2676                 nfs4_stateid *stateid,
2677                 const struct cred *cred)
2678 {
2679         return -NFS4ERR_BAD_STATEID;
2680 }
2681
2682 #if defined(CONFIG_NFS_V4_1)
2683 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2684                 nfs4_stateid *stateid,
2685                 const struct cred *cred)
2686 {
2687         int status;
2688
2689         switch (stateid->type) {
2690         default:
2691                 break;
2692         case NFS4_INVALID_STATEID_TYPE:
2693         case NFS4_SPECIAL_STATEID_TYPE:
2694                 return -NFS4ERR_BAD_STATEID;
2695         case NFS4_REVOKED_STATEID_TYPE:
2696                 goto out_free;
2697         }
2698
2699         status = nfs41_test_stateid(server, stateid, cred);
2700         switch (status) {
2701         case -NFS4ERR_EXPIRED:
2702         case -NFS4ERR_ADMIN_REVOKED:
2703         case -NFS4ERR_DELEG_REVOKED:
2704                 break;
2705         default:
2706                 return status;
2707         }
2708 out_free:
2709         /* Ack the revoked state to the server */
2710         nfs41_free_stateid(server, stateid, cred, true);
2711         return -NFS4ERR_EXPIRED;
2712 }
2713
2714 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2715 {
2716         struct nfs_server *server = NFS_SERVER(state->inode);
2717         nfs4_stateid stateid;
2718         struct nfs_delegation *delegation;
2719         const struct cred *cred = NULL;
2720         int status;
2721
2722         /* Get the delegation credential for use by test/free_stateid */
2723         rcu_read_lock();
2724         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2725         if (delegation == NULL) {
2726                 rcu_read_unlock();
2727                 nfs_state_clear_delegation(state);
2728                 return;
2729         }
2730
2731         nfs4_stateid_copy(&stateid, &delegation->stateid);
2732         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags)) {
2733                 rcu_read_unlock();
2734                 nfs_state_clear_delegation(state);
2735                 return;
2736         }
2737
2738         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2739                                 &delegation->flags)) {
2740                 rcu_read_unlock();
2741                 return;
2742         }
2743
2744         if (delegation->cred)
2745                 cred = get_cred(delegation->cred);
2746         rcu_read_unlock();
2747         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2748         trace_nfs4_test_delegation_stateid(state, NULL, status);
2749         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2750                 nfs_finish_clear_delegation_stateid(state, &stateid);
2751
2752         if (delegation->cred)
2753                 put_cred(cred);
2754 }
2755
2756 /**
2757  * nfs41_check_expired_locks - possibly free a lock stateid
2758  *
2759  * @state: NFSv4 state for an inode
2760  *
2761  * Returns NFS_OK if recovery for this stateid is now finished.
2762  * Otherwise a negative NFS4ERR value is returned.
2763  */
2764 static int nfs41_check_expired_locks(struct nfs4_state *state)
2765 {
2766         int status, ret = NFS_OK;
2767         struct nfs4_lock_state *lsp, *prev = NULL;
2768         struct nfs_server *server = NFS_SERVER(state->inode);
2769
2770         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2771                 goto out;
2772
2773         spin_lock(&state->state_lock);
2774         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2775                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2776                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2777
2778                         refcount_inc(&lsp->ls_count);
2779                         spin_unlock(&state->state_lock);
2780
2781                         nfs4_put_lock_state(prev);
2782                         prev = lsp;
2783
2784                         status = nfs41_test_and_free_expired_stateid(server,
2785                                         &lsp->ls_stateid,
2786                                         cred);
2787                         trace_nfs4_test_lock_stateid(state, lsp, status);
2788                         if (status == -NFS4ERR_EXPIRED ||
2789                             status == -NFS4ERR_BAD_STATEID) {
2790                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2791                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2792                                 if (!recover_lost_locks)
2793                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2794                         } else if (status != NFS_OK) {
2795                                 ret = status;
2796                                 nfs4_put_lock_state(prev);
2797                                 goto out;
2798                         }
2799                         spin_lock(&state->state_lock);
2800                 }
2801         }
2802         spin_unlock(&state->state_lock);
2803         nfs4_put_lock_state(prev);
2804 out:
2805         return ret;
2806 }
2807
2808 /**
2809  * nfs41_check_open_stateid - possibly free an open stateid
2810  *
2811  * @state: NFSv4 state for an inode
2812  *
2813  * Returns NFS_OK if recovery for this stateid is now finished.
2814  * Otherwise a negative NFS4ERR value is returned.
2815  */
2816 static int nfs41_check_open_stateid(struct nfs4_state *state)
2817 {
2818         struct nfs_server *server = NFS_SERVER(state->inode);
2819         nfs4_stateid *stateid = &state->open_stateid;
2820         const struct cred *cred = state->owner->so_cred;
2821         int status;
2822
2823         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2824                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2825                         if (nfs4_have_delegation(state->inode, state->state))
2826                                 return NFS_OK;
2827                         return -NFS4ERR_OPENMODE;
2828                 }
2829                 return -NFS4ERR_BAD_STATEID;
2830         }
2831         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2832         trace_nfs4_test_open_stateid(state, NULL, status);
2833         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2834                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2835                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2836                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2837                 clear_bit(NFS_OPEN_STATE, &state->flags);
2838                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2839                 return status;
2840         }
2841         if (nfs_open_stateid_recover_openmode(state))
2842                 return -NFS4ERR_OPENMODE;
2843         return NFS_OK;
2844 }
2845
2846 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2847 {
2848         int status;
2849
2850         nfs41_check_delegation_stateid(state);
2851         status = nfs41_check_expired_locks(state);
2852         if (status != NFS_OK)
2853                 return status;
2854         status = nfs41_check_open_stateid(state);
2855         if (status != NFS_OK)
2856                 status = nfs4_open_expired(sp, state);
2857         return status;
2858 }
2859 #endif
2860
2861 /*
2862  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2863  * fields corresponding to attributes that were used to store the verifier.
2864  * Make sure we clobber those fields in the later setattr call
2865  */
2866 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2867                                 struct iattr *sattr, struct nfs4_label **label)
2868 {
2869         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2870         __u32 attrset[3];
2871         unsigned ret;
2872         unsigned i;
2873
2874         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2875                 attrset[i] = opendata->o_res.attrset[i];
2876                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2877                         attrset[i] &= ~bitmask[i];
2878         }
2879
2880         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2881                 sattr->ia_valid : 0;
2882
2883         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2884                 if (sattr->ia_valid & ATTR_ATIME_SET)
2885                         ret |= ATTR_ATIME_SET;
2886                 else
2887                         ret |= ATTR_ATIME;
2888         }
2889
2890         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2891                 if (sattr->ia_valid & ATTR_MTIME_SET)
2892                         ret |= ATTR_MTIME_SET;
2893                 else
2894                         ret |= ATTR_MTIME;
2895         }
2896
2897         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2898                 *label = NULL;
2899         return ret;
2900 }
2901
2902 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2903                 fmode_t fmode,
2904                 int flags,
2905                 struct nfs_open_context *ctx)
2906 {
2907         struct nfs4_state_owner *sp = opendata->owner;
2908         struct nfs_server *server = sp->so_server;
2909         struct dentry *dentry;
2910         struct nfs4_state *state;
2911         unsigned int seq;
2912         int ret;
2913
2914         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2915
2916         ret = _nfs4_proc_open(opendata, ctx);
2917         if (ret != 0)
2918                 goto out;
2919
2920         state = _nfs4_opendata_to_nfs4_state(opendata);
2921         ret = PTR_ERR(state);
2922         if (IS_ERR(state))
2923                 goto out;
2924         ctx->state = state;
2925         if (server->caps & NFS_CAP_POSIX_LOCK)
2926                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2927         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2928                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2929
2930         dentry = opendata->dentry;
2931         if (d_really_is_negative(dentry)) {
2932                 struct dentry *alias;
2933                 d_drop(dentry);
2934                 alias = d_exact_alias(dentry, state->inode);
2935                 if (!alias)
2936                         alias = d_splice_alias(igrab(state->inode), dentry);
2937                 /* d_splice_alias() can't fail here - it's a non-directory */
2938                 if (alias) {
2939                         dput(ctx->dentry);
2940                         ctx->dentry = dentry = alias;
2941                 }
2942                 nfs_set_verifier(dentry,
2943                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2944         }
2945
2946         /* Parse layoutget results before we check for access */
2947         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
2948
2949         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2950         if (ret != 0)
2951                 goto out;
2952
2953         if (d_inode(dentry) == state->inode) {
2954                 nfs_inode_attach_open_context(ctx);
2955                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2956                         nfs4_schedule_stateid_recovery(server, state);
2957         }
2958
2959 out:
2960         if (!opendata->cancelled)
2961                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
2962         return ret;
2963 }
2964
2965 /*
2966  * Returns a referenced nfs4_state
2967  */
2968 static int _nfs4_do_open(struct inode *dir,
2969                         struct nfs_open_context *ctx,
2970                         int flags,
2971                         const struct nfs4_open_createattrs *c,
2972                         int *opened)
2973 {
2974         struct nfs4_state_owner  *sp;
2975         struct nfs4_state     *state = NULL;
2976         struct nfs_server       *server = NFS_SERVER(dir);
2977         struct nfs4_opendata *opendata;
2978         struct dentry *dentry = ctx->dentry;
2979         const struct cred *cred = ctx->cred;
2980         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2981         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2982         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2983         struct iattr *sattr = c->sattr;
2984         struct nfs4_label *label = c->label;
2985         struct nfs4_label *olabel = NULL;
2986         int status;
2987
2988         /* Protect against reboot recovery conflicts */
2989         status = -ENOMEM;
2990         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2991         if (sp == NULL) {
2992                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2993                 goto out_err;
2994         }
2995         status = nfs4_client_recover_expired_lease(server->nfs_client);
2996         if (status != 0)
2997                 goto err_put_state_owner;
2998         if (d_really_is_positive(dentry))
2999                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3000         status = -ENOMEM;
3001         if (d_really_is_positive(dentry))
3002                 claim = NFS4_OPEN_CLAIM_FH;
3003         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3004                         c, claim, GFP_KERNEL);
3005         if (opendata == NULL)
3006                 goto err_put_state_owner;
3007
3008         if (label) {
3009                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
3010                 if (IS_ERR(olabel)) {
3011                         status = PTR_ERR(olabel);
3012                         goto err_opendata_put;
3013                 }
3014         }
3015
3016         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3017                 if (!opendata->f_attr.mdsthreshold) {
3018                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3019                         if (!opendata->f_attr.mdsthreshold)
3020                                 goto err_free_label;
3021                 }
3022                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3023         }
3024         if (d_really_is_positive(dentry))
3025                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3026
3027         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
3028         if (status != 0)
3029                 goto err_free_label;
3030         state = ctx->state;
3031
3032         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3033             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3034                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3035                 /*
3036                  * send create attributes which was not set by open
3037                  * with an extra setattr.
3038                  */
3039                 if (attrs || label) {
3040                         unsigned ia_old = sattr->ia_valid;
3041
3042                         sattr->ia_valid = attrs;
3043                         nfs_fattr_init(opendata->o_res.f_attr);
3044                         status = nfs4_do_setattr(state->inode, cred,
3045                                         opendata->o_res.f_attr, sattr,
3046                                         ctx, label, olabel);
3047                         if (status == 0) {
3048                                 nfs_setattr_update_inode(state->inode, sattr,
3049                                                 opendata->o_res.f_attr);
3050                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3051                         }
3052                         sattr->ia_valid = ia_old;
3053                 }
3054         }
3055         if (opened && opendata->file_created)
3056                 *opened = 1;
3057
3058         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3059                 *ctx_th = opendata->f_attr.mdsthreshold;
3060                 opendata->f_attr.mdsthreshold = NULL;
3061         }
3062
3063         nfs4_label_free(olabel);
3064
3065         nfs4_opendata_put(opendata);
3066         nfs4_put_state_owner(sp);
3067         return 0;
3068 err_free_label:
3069         nfs4_label_free(olabel);
3070 err_opendata_put:
3071         nfs4_opendata_put(opendata);
3072 err_put_state_owner:
3073         nfs4_put_state_owner(sp);
3074 out_err:
3075         return status;
3076 }
3077
3078
3079 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3080                                         struct nfs_open_context *ctx,
3081                                         int flags,
3082                                         struct iattr *sattr,
3083                                         struct nfs4_label *label,
3084                                         int *opened)
3085 {
3086         struct nfs_server *server = NFS_SERVER(dir);
3087         struct nfs4_exception exception = {
3088                 .interruptible = true,
3089         };
3090         struct nfs4_state *res;
3091         struct nfs4_open_createattrs c = {
3092                 .label = label,
3093                 .sattr = sattr,
3094                 .verf = {
3095                         [0] = (__u32)jiffies,
3096                         [1] = (__u32)current->pid,
3097                 },
3098         };
3099         int status;
3100
3101         do {
3102                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3103                 res = ctx->state;
3104                 trace_nfs4_open_file(ctx, flags, status);
3105                 if (status == 0)
3106                         break;
3107                 /* NOTE: BAD_SEQID means the server and client disagree about the
3108                  * book-keeping w.r.t. state-changing operations
3109                  * (OPEN/CLOSE/LOCK/LOCKU...)
3110                  * It is actually a sign of a bug on the client or on the server.
3111                  *
3112                  * If we receive a BAD_SEQID error in the particular case of
3113                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3114                  * have unhashed the old state_owner for us, and that we can
3115                  * therefore safely retry using a new one. We should still warn
3116                  * the user though...
3117                  */
3118                 if (status == -NFS4ERR_BAD_SEQID) {
3119                         pr_warn_ratelimited("NFS: v4 server %s "
3120                                         " returned a bad sequence-id error!\n",
3121                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3122                         exception.retry = 1;
3123                         continue;
3124                 }
3125                 /*
3126                  * BAD_STATEID on OPEN means that the server cancelled our
3127                  * state before it received the OPEN_CONFIRM.
3128                  * Recover by retrying the request as per the discussion
3129                  * on Page 181 of RFC3530.
3130                  */
3131                 if (status == -NFS4ERR_BAD_STATEID) {
3132                         exception.retry = 1;
3133                         continue;
3134                 }
3135                 if (status == -EAGAIN) {
3136                         /* We must have found a delegation */
3137                         exception.retry = 1;
3138                         continue;
3139                 }
3140                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3141                         continue;
3142                 res = ERR_PTR(nfs4_handle_exception(server,
3143                                         status, &exception));
3144         } while (exception.retry);
3145         return res;
3146 }
3147
3148 static int _nfs4_do_setattr(struct inode *inode,
3149                             struct nfs_setattrargs *arg,
3150                             struct nfs_setattrres *res,
3151                             const struct cred *cred,
3152                             struct nfs_open_context *ctx)
3153 {
3154         struct nfs_server *server = NFS_SERVER(inode);
3155         struct rpc_message msg = {
3156                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3157                 .rpc_argp       = arg,
3158                 .rpc_resp       = res,
3159                 .rpc_cred       = cred,
3160         };
3161         const struct cred *delegation_cred = NULL;
3162         unsigned long timestamp = jiffies;
3163         bool truncate;
3164         int status;
3165
3166         nfs_fattr_init(res->fattr);
3167
3168         /* Servers should only apply open mode checks for file size changes */
3169         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3170         if (!truncate)
3171                 goto zero_stateid;
3172
3173         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3174                 /* Use that stateid */
3175         } else if (ctx != NULL) {
3176                 struct nfs_lock_context *l_ctx;
3177                 if (!nfs4_valid_open_stateid(ctx->state))
3178                         return -EBADF;
3179                 l_ctx = nfs_get_lock_context(ctx);
3180                 if (IS_ERR(l_ctx))
3181                         return PTR_ERR(l_ctx);
3182                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3183                                                 &arg->stateid, &delegation_cred);
3184                 nfs_put_lock_context(l_ctx);
3185                 if (status == -EIO)
3186                         return -EBADF;
3187         } else {
3188 zero_stateid:
3189                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3190         }
3191         if (delegation_cred)
3192                 msg.rpc_cred = delegation_cred;
3193
3194         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3195
3196         put_cred(delegation_cred);
3197         if (status == 0 && ctx != NULL)
3198                 renew_lease(server, timestamp);
3199         trace_nfs4_setattr(inode, &arg->stateid, status);
3200         return status;
3201 }
3202
3203 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3204                            struct nfs_fattr *fattr, struct iattr *sattr,
3205                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3206                            struct nfs4_label *olabel)
3207 {
3208         struct nfs_server *server = NFS_SERVER(inode);
3209         __u32 bitmask[NFS4_BITMASK_SZ];
3210         struct nfs4_state *state = ctx ? ctx->state : NULL;
3211         struct nfs_setattrargs  arg = {
3212                 .fh             = NFS_FH(inode),
3213                 .iap            = sattr,
3214                 .server         = server,
3215                 .bitmask = bitmask,
3216                 .label          = ilabel,
3217         };
3218         struct nfs_setattrres  res = {
3219                 .fattr          = fattr,
3220                 .label          = olabel,
3221                 .server         = server,
3222         };
3223         struct nfs4_exception exception = {
3224                 .state = state,
3225                 .inode = inode,
3226                 .stateid = &arg.stateid,
3227         };
3228         int err;
3229
3230         do {
3231                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3232                                 nfs4_bitmask(server, olabel),
3233                                 inode);
3234
3235                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3236                 switch (err) {
3237                 case -NFS4ERR_OPENMODE:
3238                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3239                                 pr_warn_once("NFSv4: server %s is incorrectly "
3240                                                 "applying open mode checks to "
3241                                                 "a SETATTR that is not "
3242                                                 "changing file size.\n",
3243                                                 server->nfs_client->cl_hostname);
3244                         }
3245                         if (state && !(state->state & FMODE_WRITE)) {
3246                                 err = -EBADF;
3247                                 if (sattr->ia_valid & ATTR_OPEN)
3248                                         err = -EACCES;
3249                                 goto out;
3250                         }
3251                 }
3252                 err = nfs4_handle_exception(server, err, &exception);
3253         } while (exception.retry);
3254 out:
3255         return err;
3256 }
3257
3258 static bool
3259 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3260 {
3261         if (inode == NULL || !nfs_have_layout(inode))
3262                 return false;
3263
3264         return pnfs_wait_on_layoutreturn(inode, task);
3265 }
3266
3267 struct nfs4_closedata {
3268         struct inode *inode;
3269         struct nfs4_state *state;
3270         struct nfs_closeargs arg;
3271         struct nfs_closeres res;
3272         struct {
3273                 struct nfs4_layoutreturn_args arg;
3274                 struct nfs4_layoutreturn_res res;
3275                 struct nfs4_xdr_opaque_data ld_private;
3276                 u32 roc_barrier;
3277                 bool roc;
3278         } lr;
3279         struct nfs_fattr fattr;
3280         unsigned long timestamp;
3281 };
3282
3283 static void nfs4_free_closedata(void *data)
3284 {
3285         struct nfs4_closedata *calldata = data;
3286         struct nfs4_state_owner *sp = calldata->state->owner;
3287         struct super_block *sb = calldata->state->inode->i_sb;
3288
3289         if (calldata->lr.roc)
3290                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3291                                 calldata->res.lr_ret);
3292         nfs4_put_open_state(calldata->state);
3293         nfs_free_seqid(calldata->arg.seqid);
3294         nfs4_put_state_owner(sp);
3295         nfs_sb_deactive(sb);
3296         kfree(calldata);
3297 }
3298
3299 static void nfs4_close_done(struct rpc_task *task, void *data)
3300 {
3301         struct nfs4_closedata *calldata = data;
3302         struct nfs4_state *state = calldata->state;
3303         struct nfs_server *server = NFS_SERVER(calldata->inode);
3304         nfs4_stateid *res_stateid = NULL;
3305         struct nfs4_exception exception = {
3306                 .state = state,
3307                 .inode = calldata->inode,
3308                 .stateid = &calldata->arg.stateid,
3309         };
3310
3311         dprintk("%s: begin!\n", __func__);
3312         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3313                 return;
3314         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3315
3316         /* Handle Layoutreturn errors */
3317         if (calldata->arg.lr_args && task->tk_status != 0) {
3318                 switch (calldata->res.lr_ret) {
3319                 default:
3320                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3321                         break;
3322                 case 0:
3323                         calldata->arg.lr_args = NULL;
3324                         calldata->res.lr_res = NULL;
3325                         break;
3326                 case -NFS4ERR_OLD_STATEID:
3327                         if (nfs4_layoutreturn_refresh_stateid(&calldata->arg.lr_args->stateid,
3328                                                 &calldata->arg.lr_args->range,
3329                                                 calldata->inode))
3330                                 goto lr_restart;
3331                         /* Fallthrough */
3332                 case -NFS4ERR_ADMIN_REVOKED:
3333                 case -NFS4ERR_DELEG_REVOKED:
3334                 case -NFS4ERR_EXPIRED:
3335                 case -NFS4ERR_BAD_STATEID:
3336                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3337                 case -NFS4ERR_WRONG_CRED:
3338                         calldata->arg.lr_args = NULL;
3339                         calldata->res.lr_res = NULL;
3340                         goto lr_restart;
3341                 }
3342         }
3343
3344         /* hmm. we are done with the inode, and in the process of freeing
3345          * the state_owner. we keep this around to process errors
3346          */
3347         switch (task->tk_status) {
3348                 case 0:
3349                         res_stateid = &calldata->res.stateid;
3350                         renew_lease(server, calldata->timestamp);
3351                         break;
3352                 case -NFS4ERR_ACCESS:
3353                         if (calldata->arg.bitmask != NULL) {
3354                                 calldata->arg.bitmask = NULL;
3355                                 calldata->res.fattr = NULL;
3356                                 goto out_restart;
3357
3358                         }
3359                         break;
3360                 case -NFS4ERR_OLD_STATEID:
3361                         /* Did we race with OPEN? */
3362                         if (nfs4_refresh_open_stateid(&calldata->arg.stateid,
3363                                                 state))
3364                                 goto out_restart;
3365                         goto out_release;
3366                 case -NFS4ERR_ADMIN_REVOKED:
3367                 case -NFS4ERR_STALE_STATEID:
3368                 case -NFS4ERR_EXPIRED:
3369                         nfs4_free_revoked_stateid(server,
3370                                         &calldata->arg.stateid,
3371                                         task->tk_msg.rpc_cred);
3372                         /* Fallthrough */
3373                 case -NFS4ERR_BAD_STATEID:
3374                         break;
3375                 default:
3376                         task->tk_status = nfs4_async_handle_exception(task,
3377                                         server, task->tk_status, &exception);
3378                         if (exception.retry)
3379                                 goto out_restart;
3380         }
3381         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3382                         res_stateid, calldata->arg.fmode);
3383 out_release:
3384         task->tk_status = 0;
3385         nfs_release_seqid(calldata->arg.seqid);
3386         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3387         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3388         return;
3389 lr_restart:
3390         calldata->res.lr_ret = 0;
3391 out_restart:
3392         task->tk_status = 0;
3393         rpc_restart_call_prepare(task);
3394         goto out_release;
3395 }
3396
3397 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3398 {
3399         struct nfs4_closedata *calldata = data;
3400         struct nfs4_state *state = calldata->state;
3401         struct inode *inode = calldata->inode;
3402         struct pnfs_layout_hdr *lo;
3403         bool is_rdonly, is_wronly, is_rdwr;
3404         int call_close = 0;
3405
3406         dprintk("%s: begin!\n", __func__);
3407         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3408                 goto out_wait;
3409
3410         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3411         spin_lock(&state->owner->so_lock);
3412         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3413         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3414         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3415         /* Calculate the change in open mode */
3416         calldata->arg.fmode = 0;
3417         if (state->n_rdwr == 0) {
3418                 if (state->n_rdonly == 0)
3419                         call_close |= is_rdonly;
3420                 else if (is_rdonly)
3421                         calldata->arg.fmode |= FMODE_READ;
3422                 if (state->n_wronly == 0)
3423                         call_close |= is_wronly;
3424                 else if (is_wronly)
3425                         calldata->arg.fmode |= FMODE_WRITE;
3426                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3427                         call_close |= is_rdwr;
3428         } else if (is_rdwr)
3429                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3430
3431         if (!nfs4_valid_open_stateid(state) ||
3432             !nfs4_refresh_open_stateid(&calldata->arg.stateid, state))
3433                 call_close = 0;
3434         spin_unlock(&state->owner->so_lock);
3435
3436         if (!call_close) {
3437                 /* Note: exit _without_ calling nfs4_close_done */
3438                 goto out_no_action;
3439         }
3440
3441         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3442                 nfs_release_seqid(calldata->arg.seqid);
3443                 goto out_wait;
3444         }
3445
3446         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3447         if (lo && !pnfs_layout_is_valid(lo)) {
3448                 calldata->arg.lr_args = NULL;
3449                 calldata->res.lr_res = NULL;
3450         }
3451
3452         if (calldata->arg.fmode == 0)
3453                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3454
3455         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3456                 /* Close-to-open cache consistency revalidation */
3457                 if (!nfs4_have_delegation(inode, FMODE_READ))
3458                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3459                 else
3460                         calldata->arg.bitmask = NULL;
3461         }
3462
3463         calldata->arg.share_access =
3464                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3465                                 calldata->arg.fmode, 0);
3466
3467         if (calldata->res.fattr == NULL)
3468                 calldata->arg.bitmask = NULL;
3469         else if (calldata->arg.bitmask == NULL)
3470                 calldata->res.fattr = NULL;
3471         calldata->timestamp = jiffies;
3472         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3473                                 &calldata->arg.seq_args,
3474                                 &calldata->res.seq_res,
3475                                 task) != 0)
3476                 nfs_release_seqid(calldata->arg.seqid);
3477         dprintk("%s: done!\n", __func__);
3478         return;
3479 out_no_action:
3480         task->tk_action = NULL;
3481 out_wait:
3482         nfs4_sequence_done(task, &calldata->res.seq_res);
3483 }
3484
3485 static const struct rpc_call_ops nfs4_close_ops = {
3486         .rpc_call_prepare = nfs4_close_prepare,
3487         .rpc_call_done = nfs4_close_done,
3488         .rpc_release = nfs4_free_closedata,
3489 };
3490
3491 /* 
3492  * It is possible for data to be read/written from a mem-mapped file 
3493  * after the sys_close call (which hits the vfs layer as a flush).
3494  * This means that we can't safely call nfsv4 close on a file until 
3495  * the inode is cleared. This in turn means that we are not good
3496  * NFSv4 citizens - we do not indicate to the server to update the file's 
3497  * share state even when we are done with one of the three share 
3498  * stateid's in the inode.
3499  *
3500  * NOTE: Caller must be holding the sp->so_owner semaphore!
3501  */
3502 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3503 {
3504         struct nfs_server *server = NFS_SERVER(state->inode);
3505         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3506         struct nfs4_closedata *calldata;
3507         struct nfs4_state_owner *sp = state->owner;
3508         struct rpc_task *task;
3509         struct rpc_message msg = {
3510                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3511                 .rpc_cred = state->owner->so_cred,
3512         };
3513         struct rpc_task_setup task_setup_data = {
3514                 .rpc_client = server->client,
3515                 .rpc_message = &msg,
3516                 .callback_ops = &nfs4_close_ops,
3517                 .workqueue = nfsiod_workqueue,
3518                 .flags = RPC_TASK_ASYNC,
3519         };
3520         int status = -ENOMEM;
3521
3522         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3523                 &task_setup_data.rpc_client, &msg);
3524
3525         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3526         if (calldata == NULL)
3527                 goto out;
3528         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3529         calldata->inode = state->inode;
3530         calldata->state = state;
3531         calldata->arg.fh = NFS_FH(state->inode);
3532         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3533                 goto out_free_calldata;
3534         /* Serialization for the sequence id */
3535         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3536         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3537         if (IS_ERR(calldata->arg.seqid))
3538                 goto out_free_calldata;
3539         nfs_fattr_init(&calldata->fattr);
3540         calldata->arg.fmode = 0;
3541         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3542         calldata->res.fattr = &calldata->fattr;
3543         calldata->res.seqid = calldata->arg.seqid;
3544         calldata->res.server = server;
3545         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3546         calldata->lr.roc = pnfs_roc(state->inode,
3547                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3548         if (calldata->lr.roc) {
3549                 calldata->arg.lr_args = &calldata->lr.arg;
3550                 calldata->res.lr_res = &calldata->lr.res;
3551         }
3552         nfs_sb_active(calldata->inode->i_sb);
3553
3554         msg.rpc_argp = &calldata->arg;
3555         msg.rpc_resp = &calldata->res;
3556         task_setup_data.callback_data = calldata;
3557         task = rpc_run_task(&task_setup_data);
3558         if (IS_ERR(task))
3559                 return PTR_ERR(task);
3560         status = 0;
3561         if (wait)
3562                 status = rpc_wait_for_completion_task(task);
3563         rpc_put_task(task);
3564         return status;
3565 out_free_calldata:
3566         kfree(calldata);
3567 out:
3568         nfs4_put_open_state(state);
3569         nfs4_put_state_owner(sp);
3570         return status;
3571 }
3572
3573 static struct inode *
3574 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3575                 int open_flags, struct iattr *attr, int *opened)
3576 {
3577         struct nfs4_state *state;
3578         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3579
3580         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3581
3582         /* Protect against concurrent sillydeletes */
3583         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3584
3585         nfs4_label_release_security(label);
3586
3587         if (IS_ERR(state))
3588                 return ERR_CAST(state);
3589         return state->inode;
3590 }
3591
3592 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3593 {
3594         if (ctx->state == NULL)
3595                 return;
3596         if (is_sync)
3597                 nfs4_close_sync(ctx->state, ctx->mode);
3598         else
3599                 nfs4_close_state(ctx->state, ctx->mode);
3600 }
3601
3602 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3603 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3604 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3605
3606 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3607 {
3608         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3609         struct nfs4_server_caps_arg args = {
3610                 .fhandle = fhandle,
3611                 .bitmask = bitmask,
3612         };
3613         struct nfs4_server_caps_res res = {};
3614         struct rpc_message msg = {
3615                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3616                 .rpc_argp = &args,
3617                 .rpc_resp = &res,
3618         };
3619         int status;
3620         int i;
3621
3622         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3623                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3624                      FATTR4_WORD0_LINK_SUPPORT |
3625                      FATTR4_WORD0_SYMLINK_SUPPORT |
3626                      FATTR4_WORD0_ACLSUPPORT;
3627         if (minorversion)
3628                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3629
3630         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3631         if (status == 0) {
3632                 /* Sanity check the server answers */
3633                 switch (minorversion) {
3634                 case 0:
3635                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3636                         res.attr_bitmask[2] = 0;
3637                         break;
3638                 case 1:
3639                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3640                         break;
3641                 case 2:
3642                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3643                 }
3644                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3645                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3646                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3647                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3648                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3649                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3650                                 NFS_CAP_SECURITY_LABEL);
3651                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3652                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3653                         server->caps |= NFS_CAP_ACLS;
3654                 if (res.has_links != 0)
3655                         server->caps |= NFS_CAP_HARDLINKS;
3656                 if (res.has_symlinks != 0)
3657                         server->caps |= NFS_CAP_SYMLINKS;
3658                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3659                         server->caps |= NFS_CAP_FILEID;
3660                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3661                         server->caps |= NFS_CAP_MODE;
3662                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3663                         server->caps |= NFS_CAP_NLINK;
3664                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3665                         server->caps |= NFS_CAP_OWNER;
3666                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3667                         server->caps |= NFS_CAP_OWNER_GROUP;
3668                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3669                         server->caps |= NFS_CAP_ATIME;
3670                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3671                         server->caps |= NFS_CAP_CTIME;
3672                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3673                         server->caps |= NFS_CAP_MTIME;
3674 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3675                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3676                         server->caps |= NFS_CAP_SECURITY_LABEL;
3677 #endif
3678                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3679                                 sizeof(server->attr_bitmask));
3680                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3681
3682                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3683                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3684                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3685                 server->cache_consistency_bitmask[2] = 0;
3686
3687                 /* Avoid a regression due to buggy server */
3688                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3689                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3690                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3691                         sizeof(server->exclcreat_bitmask));
3692
3693                 server->acl_bitmask = res.acl_bitmask;
3694                 server->fh_expire_type = res.fh_expire_type;
3695         }
3696
3697         return status;
3698 }
3699
3700 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3701 {
3702         struct nfs4_exception exception = {
3703                 .interruptible = true,
3704         };
3705         int err;
3706         do {
3707                 err = nfs4_handle_exception(server,
3708                                 _nfs4_server_capabilities(server, fhandle),
3709                                 &exception);
3710         } while (exception.retry);
3711         return err;
3712 }
3713
3714 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3715                 struct nfs_fsinfo *info)
3716 {
3717         u32 bitmask[3];
3718         struct nfs4_lookup_root_arg args = {
3719                 .bitmask = bitmask,
3720         };
3721         struct nfs4_lookup_res res = {
3722                 .server = server,
3723                 .fattr = info->fattr,
3724                 .fh = fhandle,
3725         };
3726         struct rpc_message msg = {
3727                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3728                 .rpc_argp = &args,
3729                 .rpc_resp = &res,
3730         };
3731
3732         bitmask[0] = nfs4_fattr_bitmap[0];
3733         bitmask[1] = nfs4_fattr_bitmap[1];
3734         /*
3735          * Process the label in the upcoming getfattr
3736          */
3737         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3738
3739         nfs_fattr_init(info->fattr);
3740         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3741 }
3742
3743 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3744                 struct nfs_fsinfo *info)
3745 {
3746         struct nfs4_exception exception = {
3747                 .interruptible = true,
3748         };
3749         int err;
3750         do {
3751                 err = _nfs4_lookup_root(server, fhandle, info);
3752                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3753                 switch (err) {
3754                 case 0:
3755                 case -NFS4ERR_WRONGSEC:
3756                         goto out;
3757                 default:
3758                         err = nfs4_handle_exception(server, err, &exception);
3759                 }
3760         } while (exception.retry);
3761 out:
3762         return err;
3763 }
3764
3765 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3766                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3767 {
3768         struct rpc_auth_create_args auth_args = {
3769                 .pseudoflavor = flavor,
3770         };
3771         struct rpc_auth *auth;
3772
3773         auth = rpcauth_create(&auth_args, server->client);
3774         if (IS_ERR(auth))
3775                 return -EACCES;
3776         return nfs4_lookup_root(server, fhandle, info);
3777 }
3778
3779 /*
3780  * Retry pseudoroot lookup with various security flavors.  We do this when:
3781  *
3782  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3783  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3784  *
3785  * Returns zero on success, or a negative NFS4ERR value, or a
3786  * negative errno value.
3787  */
3788 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3789                               struct nfs_fsinfo *info)
3790 {
3791         /* Per 3530bis 15.33.5 */
3792         static const rpc_authflavor_t flav_array[] = {
3793                 RPC_AUTH_GSS_KRB5P,
3794                 RPC_AUTH_GSS_KRB5I,
3795                 RPC_AUTH_GSS_KRB5,
3796                 RPC_AUTH_UNIX,                  /* courtesy */
3797                 RPC_AUTH_NULL,
3798         };
3799         int status = -EPERM;
3800         size_t i;
3801
3802         if (server->auth_info.flavor_len > 0) {
3803                 /* try each flavor specified by user */
3804                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3805                         status = nfs4_lookup_root_sec(server, fhandle, info,
3806                                                 server->auth_info.flavors[i]);
3807                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3808                                 continue;
3809                         break;
3810                 }
3811         } else {
3812                 /* no flavors specified by user, try default list */
3813                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3814                         status = nfs4_lookup_root_sec(server, fhandle, info,
3815                                                       flav_array[i]);
3816                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3817                                 continue;
3818                         break;
3819                 }
3820         }
3821
3822         /*
3823          * -EACCES could mean that the user doesn't have correct permissions
3824          * to access the mount.  It could also mean that we tried to mount
3825          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3826          * existing mount programs don't handle -EACCES very well so it should
3827          * be mapped to -EPERM instead.
3828          */
3829         if (status == -EACCES)
3830                 status = -EPERM;
3831         return status;
3832 }
3833
3834 /**
3835  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3836  * @server: initialized nfs_server handle
3837  * @fhandle: we fill in the pseudo-fs root file handle
3838  * @info: we fill in an FSINFO struct
3839  * @auth_probe: probe the auth flavours
3840  *
3841  * Returns zero on success, or a negative errno.
3842  */
3843 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3844                          struct nfs_fsinfo *info,
3845                          bool auth_probe)
3846 {
3847         int status = 0;
3848
3849         if (!auth_probe)
3850                 status = nfs4_lookup_root(server, fhandle, info);
3851
3852         if (auth_probe || status == NFS4ERR_WRONGSEC)
3853                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3854                                 fhandle, info);
3855
3856         if (status == 0)
3857                 status = nfs4_server_capabilities(server, fhandle);
3858         if (status == 0)
3859                 status = nfs4_do_fsinfo(server, fhandle, info);
3860
3861         return nfs4_map_errors(status);
3862 }
3863
3864 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3865                               struct nfs_fsinfo *info)
3866 {
3867         int error;
3868         struct nfs_fattr *fattr = info->fattr;
3869         struct nfs4_label *label = NULL;
3870
3871         error = nfs4_server_capabilities(server, mntfh);
3872         if (error < 0) {
3873                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3874                 return error;
3875         }
3876
3877         label = nfs4_label_alloc(server, GFP_KERNEL);
3878         if (IS_ERR(label))
3879                 return PTR_ERR(label);
3880
3881         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
3882         if (error < 0) {
3883                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3884                 goto err_free_label;
3885         }
3886
3887         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3888             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3889                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3890
3891 err_free_label:
3892         nfs4_label_free(label);
3893
3894         return error;
3895 }
3896
3897 /*
3898  * Get locations and (maybe) other attributes of a referral.
3899  * Note that we'll actually follow the referral later when
3900  * we detect fsid mismatch in inode revalidation
3901  */
3902 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3903                              const struct qstr *name, struct nfs_fattr *fattr,
3904                              struct nfs_fh *fhandle)
3905 {
3906         int status = -ENOMEM;
3907         struct page *page = NULL;
3908         struct nfs4_fs_locations *locations = NULL;
3909
3910         page = alloc_page(GFP_KERNEL);
3911         if (page == NULL)
3912                 goto out;
3913         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3914         if (locations == NULL)
3915                 goto out;
3916
3917         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3918         if (status != 0)
3919                 goto out;
3920
3921         /*
3922          * If the fsid didn't change, this is a migration event, not a
3923          * referral.  Cause us to drop into the exception handler, which
3924          * will kick off migration recovery.
3925          */
3926         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3927                 dprintk("%s: server did not return a different fsid for"
3928                         " a referral at %s\n", __func__, name->name);
3929                 status = -NFS4ERR_MOVED;
3930                 goto out;
3931         }
3932         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3933         nfs_fixup_referral_attributes(&locations->fattr);
3934
3935         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3936         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3937         memset(fhandle, 0, sizeof(struct nfs_fh));
3938 out:
3939         if (page)
3940                 __free_page(page);
3941         kfree(locations);
3942         return status;
3943 }
3944
3945 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3946                                 struct nfs_fattr *fattr, struct nfs4_label *label,
3947                                 struct inode *inode)
3948 {
3949         __u32 bitmask[NFS4_BITMASK_SZ];
3950         struct nfs4_getattr_arg args = {
3951                 .fh = fhandle,
3952                 .bitmask = bitmask,
3953         };
3954         struct nfs4_getattr_res res = {
3955                 .fattr = fattr,
3956                 .label = label,
3957                 .server = server,
3958         };
3959         struct rpc_message msg = {
3960                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3961                 .rpc_argp = &args,
3962                 .rpc_resp = &res,
3963         };
3964
3965         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
3966
3967         nfs_fattr_init(fattr);
3968         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3969 }
3970
3971 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3972                                 struct nfs_fattr *fattr, struct nfs4_label *label,
3973                                 struct inode *inode)
3974 {
3975         struct nfs4_exception exception = {
3976                 .interruptible = true,
3977         };
3978         int err;
3979         do {
3980                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
3981                 trace_nfs4_getattr(server, fhandle, fattr, err);
3982                 err = nfs4_handle_exception(server, err,
3983                                 &exception);
3984         } while (exception.retry);
3985         return err;
3986 }
3987
3988 /* 
3989  * The file is not closed if it is opened due to the a request to change
3990  * the size of the file. The open call will not be needed once the
3991  * VFS layer lookup-intents are implemented.
3992  *
3993  * Close is called when the inode is destroyed.
3994  * If we haven't opened the file for O_WRONLY, we
3995  * need to in the size_change case to obtain a stateid.
3996  *
3997  * Got race?
3998  * Because OPEN is always done by name in nfsv4, it is
3999  * possible that we opened a different file by the same
4000  * name.  We can recognize this race condition, but we
4001  * can't do anything about it besides returning an error.
4002  *
4003  * This will be fixed with VFS changes (lookup-intent).
4004  */
4005 static int
4006 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4007                   struct iattr *sattr)
4008 {
4009         struct inode *inode = d_inode(dentry);
4010         const struct cred *cred = NULL;
4011         struct nfs_open_context *ctx = NULL;
4012         struct nfs4_label *label = NULL;
4013         int status;
4014
4015         if (pnfs_ld_layoutret_on_setattr(inode) &&
4016             sattr->ia_valid & ATTR_SIZE &&
4017             sattr->ia_size < i_size_read(inode))
4018                 pnfs_commit_and_return_layout(inode);
4019
4020         nfs_fattr_init(fattr);
4021         
4022         /* Deal with open(O_TRUNC) */
4023         if (sattr->ia_valid & ATTR_OPEN)
4024                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4025
4026         /* Optimization: if the end result is no change, don't RPC */
4027         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4028                 return 0;
4029
4030         /* Search for an existing open(O_WRITE) file */
4031         if (sattr->ia_valid & ATTR_FILE) {
4032
4033                 ctx = nfs_file_open_context(sattr->ia_file);
4034                 if (ctx)
4035                         cred = ctx->cred;
4036         }
4037
4038         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4039         if (IS_ERR(label))
4040                 return PTR_ERR(label);
4041
4042         /* Return any delegations if we're going to change ACLs */
4043         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4044                 nfs4_inode_make_writeable(inode);
4045
4046         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4047         if (status == 0) {
4048                 nfs_setattr_update_inode(inode, sattr, fattr);
4049                 nfs_setsecurity(inode, fattr, label);
4050         }
4051         nfs4_label_free(label);
4052         return status;
4053 }
4054
4055 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4056                 const struct qstr *name, struct nfs_fh *fhandle,
4057                 struct nfs_fattr *fattr, struct nfs4_label *label)
4058 {
4059         struct nfs_server *server = NFS_SERVER(dir);
4060         int                    status;
4061         struct nfs4_lookup_arg args = {
4062                 .bitmask = server->attr_bitmask,
4063                 .dir_fh = NFS_FH(dir),
4064                 .name = name,
4065         };
4066         struct nfs4_lookup_res res = {
4067                 .server = server,
4068                 .fattr = fattr,
4069                 .label = label,
4070                 .fh = fhandle,
4071         };
4072         struct rpc_message msg = {
4073                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4074                 .rpc_argp = &args,
4075                 .rpc_resp = &res,
4076         };
4077
4078         args.bitmask = nfs4_bitmask(server, label);
4079
4080         nfs_fattr_init(fattr);
4081
4082         dprintk("NFS call  lookup %s\n", name->name);
4083         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
4084         dprintk("NFS reply lookup: %d\n", status);
4085         return status;
4086 }
4087
4088 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4089 {
4090         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4091                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4092         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4093         fattr->nlink = 2;
4094 }
4095
4096 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4097                                    const struct qstr *name, struct nfs_fh *fhandle,
4098                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4099 {
4100         struct nfs4_exception exception = {
4101                 .interruptible = true,
4102         };
4103         struct rpc_clnt *client = *clnt;
4104         int err;
4105         do {
4106                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
4107                 trace_nfs4_lookup(dir, name, err);
4108                 switch (err) {
4109                 case -NFS4ERR_BADNAME:
4110                         err = -ENOENT;
4111                         goto out;
4112                 case -NFS4ERR_MOVED:
4113                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4114                         if (err == -NFS4ERR_MOVED)
4115                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4116                         goto out;
4117                 case -NFS4ERR_WRONGSEC:
4118                         err = -EPERM;
4119                         if (client != *clnt)
4120                                 goto out;
4121                         client = nfs4_negotiate_security(client, dir, name);
4122                         if (IS_ERR(client))
4123                                 return PTR_ERR(client);
4124
4125                         exception.retry = 1;
4126                         break;
4127                 default:
4128                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4129                 }
4130         } while (exception.retry);
4131
4132 out:
4133         if (err == 0)
4134                 *clnt = client;
4135         else if (client != *clnt)
4136                 rpc_shutdown_client(client);
4137
4138         return err;
4139 }
4140
4141 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
4142                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4143                             struct nfs4_label *label)
4144 {
4145         int status;
4146         struct rpc_clnt *client = NFS_CLIENT(dir);
4147
4148         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
4149         if (client != NFS_CLIENT(dir)) {
4150                 rpc_shutdown_client(client);
4151                 nfs_fixup_secinfo_attributes(fattr);
4152         }
4153         return status;
4154 }
4155
4156 struct rpc_clnt *
4157 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
4158                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4159 {
4160         struct rpc_clnt *client = NFS_CLIENT(dir);
4161         int status;
4162
4163         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
4164         if (status < 0)
4165                 return ERR_PTR(status);
4166         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4167 }
4168
4169 static int _nfs4_proc_lookupp(struct inode *inode,
4170                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4171                 struct nfs4_label *label)
4172 {
4173         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4174         struct nfs_server *server = NFS_SERVER(inode);
4175         int                    status;
4176         struct nfs4_lookupp_arg args = {
4177                 .bitmask = server->attr_bitmask,
4178                 .fh = NFS_FH(inode),
4179         };
4180         struct nfs4_lookupp_res res = {
4181                 .server = server,
4182                 .fattr = fattr,
4183                 .label = label,
4184                 .fh = fhandle,
4185         };
4186         struct rpc_message msg = {
4187                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4188                 .rpc_argp = &args,
4189                 .rpc_resp = &res,
4190         };
4191
4192         args.bitmask = nfs4_bitmask(server, label);
4193
4194         nfs_fattr_init(fattr);
4195
4196         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4197         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4198                                 &res.seq_res, 0);
4199         dprintk("NFS reply lookupp: %d\n", status);
4200         return status;
4201 }
4202
4203 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4204                              struct nfs_fattr *fattr, struct nfs4_label *label)
4205 {
4206         struct nfs4_exception exception = {
4207                 .interruptible = true,
4208         };
4209         int err;
4210         do {
4211                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4212                 trace_nfs4_lookupp(inode, err);
4213                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4214                                 &exception);
4215         } while (exception.retry);
4216         return err;
4217 }
4218
4219 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4220 {
4221         struct nfs_server *server = NFS_SERVER(inode);
4222         struct nfs4_accessargs args = {
4223                 .fh = NFS_FH(inode),
4224                 .access = entry->mask,
4225         };
4226         struct nfs4_accessres res = {
4227                 .server = server,
4228         };
4229         struct rpc_message msg = {
4230                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4231                 .rpc_argp = &args,
4232                 .rpc_resp = &res,
4233                 .rpc_cred = entry->cred,
4234         };
4235         int status = 0;
4236
4237         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4238                 res.fattr = nfs_alloc_fattr();
4239                 if (res.fattr == NULL)
4240                         return -ENOMEM;
4241                 args.bitmask = server->cache_consistency_bitmask;
4242         }
4243         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4244         if (!status) {
4245                 nfs_access_set_mask(entry, res.access);
4246                 if (res.fattr)
4247                         nfs_refresh_inode(inode, res.fattr);
4248         }
4249         nfs_free_fattr(res.fattr);
4250         return status;
4251 }
4252
4253 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4254 {
4255         struct nfs4_exception exception = {
4256                 .interruptible = true,
4257         };
4258         int err;
4259         do {
4260                 err = _nfs4_proc_access(inode, entry);
4261                 trace_nfs4_access(inode, err);
4262                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4263                                 &exception);
4264         } while (exception.retry);
4265         return err;
4266 }
4267
4268 /*
4269  * TODO: For the time being, we don't try to get any attributes
4270  * along with any of the zero-copy operations READ, READDIR,
4271  * READLINK, WRITE.
4272  *
4273  * In the case of the first three, we want to put the GETATTR
4274  * after the read-type operation -- this is because it is hard
4275  * to predict the length of a GETATTR response in v4, and thus
4276  * align the READ data correctly.  This means that the GETATTR
4277  * may end up partially falling into the page cache, and we should
4278  * shift it into the 'tail' of the xdr_buf before processing.
4279  * To do this efficiently, we need to know the total length
4280  * of data received, which doesn't seem to be available outside
4281  * of the RPC layer.
4282  *
4283  * In the case of WRITE, we also want to put the GETATTR after
4284  * the operation -- in this case because we want to make sure
4285  * we get the post-operation mtime and size.
4286  *
4287  * Both of these changes to the XDR layer would in fact be quite
4288  * minor, but I decided to leave them for a subsequent patch.
4289  */
4290 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4291                 unsigned int pgbase, unsigned int pglen)
4292 {
4293         struct nfs4_readlink args = {
4294                 .fh       = NFS_FH(inode),
4295                 .pgbase   = pgbase,
4296                 .pglen    = pglen,
4297                 .pages    = &page,
4298         };
4299         struct nfs4_readlink_res res;
4300         struct rpc_message msg = {
4301                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4302                 .rpc_argp = &args,
4303                 .rpc_resp = &res,
4304         };
4305
4306         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4307 }
4308
4309 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4310                 unsigned int pgbase, unsigned int pglen)
4311 {
4312         struct nfs4_exception exception = {
4313                 .interruptible = true,
4314         };
4315         int err;
4316         do {
4317                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4318                 trace_nfs4_readlink(inode, err);
4319                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4320                                 &exception);
4321         } while (exception.retry);
4322         return err;
4323 }
4324
4325 /*
4326  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4327  */
4328 static int
4329 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4330                  int flags)
4331 {
4332         struct nfs_server *server = NFS_SERVER(dir);
4333         struct nfs4_label l, *ilabel = NULL;
4334         struct nfs_open_context *ctx;
4335         struct nfs4_state *state;
4336         int status = 0;
4337
4338         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4339         if (IS_ERR(ctx))
4340                 return PTR_ERR(ctx);
4341
4342         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4343
4344         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4345                 sattr->ia_mode &= ~current_umask();
4346         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4347         if (IS_ERR(state)) {
4348                 status = PTR_ERR(state);
4349                 goto out;
4350         }
4351 out:
4352         nfs4_label_release_security(ilabel);
4353         put_nfs_open_context(ctx);
4354         return status;
4355 }
4356
4357 static int
4358 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4359 {
4360         struct nfs_server *server = NFS_SERVER(dir);
4361         struct nfs_removeargs args = {
4362                 .fh = NFS_FH(dir),
4363                 .name = *name,
4364         };
4365         struct nfs_removeres res = {
4366                 .server = server,
4367         };
4368         struct rpc_message msg = {
4369                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4370                 .rpc_argp = &args,
4371                 .rpc_resp = &res,
4372         };
4373         unsigned long timestamp = jiffies;
4374         int status;
4375
4376         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4377         if (status == 0) {
4378                 spin_lock(&dir->i_lock);
4379                 update_changeattr_locked(dir, &res.cinfo, timestamp, 0);
4380                 /* Removing a directory decrements nlink in the parent */
4381                 if (ftype == NF4DIR && dir->i_nlink > 2)
4382                         nfs4_dec_nlink_locked(dir);
4383                 spin_unlock(&dir->i_lock);
4384         }
4385         return status;
4386 }
4387
4388 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4389 {
4390         struct nfs4_exception exception = {
4391                 .interruptible = true,
4392         };
4393         struct inode *inode = d_inode(dentry);
4394         int err;
4395
4396         if (inode) {
4397                 if (inode->i_nlink == 1)
4398                         nfs4_inode_return_delegation(inode);
4399                 else
4400                         nfs4_inode_make_writeable(inode);
4401         }
4402         do {
4403                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4404                 trace_nfs4_remove(dir, &dentry->d_name, err);
4405                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4406                                 &exception);
4407         } while (exception.retry);
4408         return err;
4409 }
4410
4411 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4412 {
4413         struct nfs4_exception exception = {
4414                 .interruptible = true,
4415         };
4416         int err;
4417
4418         do {
4419                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4420                 trace_nfs4_remove(dir, name, err);
4421                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4422                                 &exception);
4423         } while (exception.retry);
4424         return err;
4425 }
4426
4427 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4428                 struct dentry *dentry,
4429                 struct inode *inode)
4430 {
4431         struct nfs_removeargs *args = msg->rpc_argp;
4432         struct nfs_removeres *res = msg->rpc_resp;
4433
4434         res->server = NFS_SB(dentry->d_sb);
4435         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4436         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4437
4438         nfs_fattr_init(res->dir_attr);
4439
4440         if (inode)
4441                 nfs4_inode_return_delegation(inode);
4442 }
4443
4444 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4445 {
4446         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4447                         &data->args.seq_args,
4448                         &data->res.seq_res,
4449                         task);
4450 }
4451
4452 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4453 {
4454         struct nfs_unlinkdata *data = task->tk_calldata;
4455         struct nfs_removeres *res = &data->res;
4456
4457         if (!nfs4_sequence_done(task, &res->seq_res))
4458                 return 0;
4459         if (nfs4_async_handle_error(task, res->server, NULL,
4460                                     &data->timeout) == -EAGAIN)
4461                 return 0;
4462         if (task->tk_status == 0)
4463                 update_changeattr(dir, &res->cinfo,
4464                                 res->dir_attr->time_start, 0);
4465         return 1;
4466 }
4467
4468 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4469                 struct dentry *old_dentry,
4470                 struct dentry *new_dentry)
4471 {
4472         struct nfs_renameargs *arg = msg->rpc_argp;
4473         struct nfs_renameres *res = msg->rpc_resp;
4474         struct inode *old_inode = d_inode(old_dentry);
4475         struct inode *new_inode = d_inode(new_dentry);
4476
4477         if (old_inode)
4478                 nfs4_inode_make_writeable(old_inode);
4479         if (new_inode)
4480                 nfs4_inode_return_delegation(new_inode);
4481         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4482         res->server = NFS_SB(old_dentry->d_sb);
4483         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4484 }
4485
4486 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4487 {
4488         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4489                         &data->args.seq_args,
4490                         &data->res.seq_res,
4491                         task);
4492 }
4493
4494 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4495                                  struct inode *new_dir)
4496 {
4497         struct nfs_renamedata *data = task->tk_calldata;
4498         struct nfs_renameres *res = &data->res;
4499
4500         if (!nfs4_sequence_done(task, &res->seq_res))
4501                 return 0;
4502         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4503                 return 0;
4504
4505         if (task->tk_status == 0) {
4506                 if (new_dir != old_dir) {
4507                         /* Note: If we moved a directory, nlink will change */
4508                         update_changeattr(old_dir, &res->old_cinfo,
4509                                         res->old_fattr->time_start,
4510                                         NFS_INO_INVALID_OTHER);
4511                         update_changeattr(new_dir, &res->new_cinfo,
4512                                         res->new_fattr->time_start,
4513                                         NFS_INO_INVALID_OTHER);
4514                 } else
4515                         update_changeattr(old_dir, &res->old_cinfo,
4516                                         res->old_fattr->time_start,
4517                                         0);
4518         }
4519         return 1;
4520 }
4521
4522 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4523 {
4524         struct nfs_server *server = NFS_SERVER(inode);
4525         __u32 bitmask[NFS4_BITMASK_SZ];
4526         struct nfs4_link_arg arg = {
4527                 .fh     = NFS_FH(inode),
4528                 .dir_fh = NFS_FH(dir),
4529                 .name   = name,
4530                 .bitmask = bitmask,
4531         };
4532         struct nfs4_link_res res = {
4533                 .server = server,
4534                 .label = NULL,
4535         };
4536         struct rpc_message msg = {
4537                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4538                 .rpc_argp = &arg,
4539                 .rpc_resp = &res,
4540         };
4541         int status = -ENOMEM;
4542
4543         res.fattr = nfs_alloc_fattr();
4544         if (res.fattr == NULL)
4545                 goto out;
4546
4547         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4548         if (IS_ERR(res.label)) {
4549                 status = PTR_ERR(res.label);
4550                 goto out;
4551         }
4552
4553         nfs4_inode_make_writeable(inode);
4554         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4555
4556         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4557         if (!status) {
4558                 update_changeattr(dir, &res.cinfo, res.fattr->time_start, 0);
4559                 status = nfs_post_op_update_inode(inode, res.fattr);
4560                 if (!status)
4561                         nfs_setsecurity(inode, res.fattr, res.label);
4562         }
4563
4564
4565         nfs4_label_free(res.label);
4566
4567 out:
4568         nfs_free_fattr(res.fattr);
4569         return status;
4570 }
4571
4572 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4573 {
4574         struct nfs4_exception exception = {
4575                 .interruptible = true,
4576         };
4577         int err;
4578         do {
4579                 err = nfs4_handle_exception(NFS_SERVER(inode),
4580                                 _nfs4_proc_link(inode, dir, name),
4581                                 &exception);
4582         } while (exception.retry);
4583         return err;
4584 }
4585
4586 struct nfs4_createdata {
4587         struct rpc_message msg;
4588         struct nfs4_create_arg arg;
4589         struct nfs4_create_res res;
4590         struct nfs_fh fh;
4591         struct nfs_fattr fattr;
4592         struct nfs4_label *label;
4593 };
4594
4595 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4596                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4597 {
4598         struct nfs4_createdata *data;
4599
4600         data = kzalloc(sizeof(*data), GFP_KERNEL);
4601         if (data != NULL) {
4602                 struct nfs_server *server = NFS_SERVER(dir);
4603
4604                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4605                 if (IS_ERR(data->label))
4606                         goto out_free;
4607
4608                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4609                 data->msg.rpc_argp = &data->arg;
4610                 data->msg.rpc_resp = &data->res;
4611                 data->arg.dir_fh = NFS_FH(dir);
4612                 data->arg.server = server;
4613                 data->arg.name = name;
4614                 data->arg.attrs = sattr;
4615                 data->arg.ftype = ftype;
4616                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4617                 data->arg.umask = current_umask();
4618                 data->res.server = server;
4619                 data->res.fh = &data->fh;
4620                 data->res.fattr = &data->fattr;
4621                 data->res.label = data->label;
4622                 nfs_fattr_init(data->res.fattr);
4623         }
4624         return data;
4625 out_free:
4626         kfree(data);
4627         return NULL;
4628 }
4629
4630 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4631 {
4632         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4633                                     &data->arg.seq_args, &data->res.seq_res, 1);
4634         if (status == 0) {
4635                 spin_lock(&dir->i_lock);
4636                 update_changeattr_locked(dir, &data->res.dir_cinfo,
4637                                 data->res.fattr->time_start, 0);
4638                 /* Creating a directory bumps nlink in the parent */
4639                 if (data->arg.ftype == NF4DIR)
4640                         nfs4_inc_nlink_locked(dir);
4641                 spin_unlock(&dir->i_lock);
4642                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4643         }
4644         return status;
4645 }
4646
4647 static void nfs4_free_createdata(struct nfs4_createdata *data)
4648 {
4649         nfs4_label_free(data->label);
4650         kfree(data);
4651 }
4652
4653 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4654                 struct page *page, unsigned int len, struct iattr *sattr,
4655                 struct nfs4_label *label)
4656 {
4657         struct nfs4_createdata *data;
4658         int status = -ENAMETOOLONG;
4659
4660         if (len > NFS4_MAXPATHLEN)
4661                 goto out;
4662
4663         status = -ENOMEM;
4664         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4665         if (data == NULL)
4666                 goto out;
4667
4668         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4669         data->arg.u.symlink.pages = &page;
4670         data->arg.u.symlink.len = len;
4671         data->arg.label = label;
4672         
4673         status = nfs4_do_create(dir, dentry, data);
4674
4675         nfs4_free_createdata(data);
4676 out:
4677         return status;
4678 }
4679
4680 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4681                 struct page *page, unsigned int len, struct iattr *sattr)
4682 {
4683         struct nfs4_exception exception = {
4684                 .interruptible = true,
4685         };
4686         struct nfs4_label l, *label = NULL;
4687         int err;
4688
4689         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4690
4691         do {
4692                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4693                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4694                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4695                                 &exception);
4696         } while (exception.retry);
4697
4698         nfs4_label_release_security(label);
4699         return err;
4700 }
4701
4702 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4703                 struct iattr *sattr, struct nfs4_label *label)
4704 {
4705         struct nfs4_createdata *data;
4706         int status = -ENOMEM;
4707
4708         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4709         if (data == NULL)
4710                 goto out;
4711
4712         data->arg.label = label;
4713         status = nfs4_do_create(dir, dentry, data);
4714
4715         nfs4_free_createdata(data);
4716 out:
4717         return status;
4718 }
4719
4720 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4721                 struct iattr *sattr)
4722 {
4723         struct nfs_server *server = NFS_SERVER(dir);
4724         struct nfs4_exception exception = {
4725                 .interruptible = true,
4726         };
4727         struct nfs4_label l, *label = NULL;
4728         int err;
4729
4730         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4731
4732         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4733                 sattr->ia_mode &= ~current_umask();
4734         do {
4735                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4736                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4737                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4738                                 &exception);
4739         } while (exception.retry);
4740         nfs4_label_release_security(label);
4741
4742         return err;
4743 }
4744
4745 static int _nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4746                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4747 {
4748         struct inode            *dir = d_inode(dentry);
4749         struct nfs4_readdir_arg args = {
4750                 .fh = NFS_FH(dir),
4751                 .pages = pages,
4752                 .pgbase = 0,
4753                 .count = count,
4754                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4755                 .plus = plus,
4756         };
4757         struct nfs4_readdir_res res;
4758         struct rpc_message msg = {
4759                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4760                 .rpc_argp = &args,
4761                 .rpc_resp = &res,
4762                 .rpc_cred = cred,
4763         };
4764         int                     status;
4765
4766         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4767                         dentry,
4768                         (unsigned long long)cookie);
4769         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4770         res.pgbase = args.pgbase;
4771         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4772         if (status >= 0) {
4773                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4774                 status += args.pgbase;
4775         }
4776
4777         nfs_invalidate_atime(dir);
4778
4779         dprintk("%s: returns %d\n", __func__, status);
4780         return status;
4781 }
4782
4783 static int nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4784                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4785 {
4786         struct nfs4_exception exception = {
4787                 .interruptible = true,
4788         };
4789         int err;
4790         do {
4791                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4792                                 pages, count, plus);
4793                 trace_nfs4_readdir(d_inode(dentry), err);
4794                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4795                                 &exception);
4796         } while (exception.retry);
4797         return err;
4798 }
4799
4800 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4801                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4802 {
4803         struct nfs4_createdata *data;
4804         int mode = sattr->ia_mode;
4805         int status = -ENOMEM;
4806
4807         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4808         if (data == NULL)
4809                 goto out;
4810
4811         if (S_ISFIFO(mode))
4812                 data->arg.ftype = NF4FIFO;
4813         else if (S_ISBLK(mode)) {
4814                 data->arg.ftype = NF4BLK;
4815                 data->arg.u.device.specdata1 = MAJOR(rdev);
4816                 data->arg.u.device.specdata2 = MINOR(rdev);
4817         }
4818         else if (S_ISCHR(mode)) {
4819                 data->arg.ftype = NF4CHR;
4820                 data->arg.u.device.specdata1 = MAJOR(rdev);
4821                 data->arg.u.device.specdata2 = MINOR(rdev);
4822         } else if (!S_ISSOCK(mode)) {
4823                 status = -EINVAL;
4824                 goto out_free;
4825         }
4826
4827         data->arg.label = label;
4828         status = nfs4_do_create(dir, dentry, data);
4829 out_free:
4830         nfs4_free_createdata(data);
4831 out:
4832         return status;
4833 }
4834
4835 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4836                 struct iattr *sattr, dev_t rdev)
4837 {
4838         struct nfs_server *server = NFS_SERVER(dir);
4839         struct nfs4_exception exception = {
4840                 .interruptible = true,
4841         };
4842         struct nfs4_label l, *label = NULL;
4843         int err;
4844
4845         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4846
4847         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4848                 sattr->ia_mode &= ~current_umask();
4849         do {
4850                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4851                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4852                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4853                                 &exception);
4854         } while (exception.retry);
4855
4856         nfs4_label_release_security(label);
4857
4858         return err;
4859 }
4860
4861 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4862                  struct nfs_fsstat *fsstat)
4863 {
4864         struct nfs4_statfs_arg args = {
4865                 .fh = fhandle,
4866                 .bitmask = server->attr_bitmask,
4867         };
4868         struct nfs4_statfs_res res = {
4869                 .fsstat = fsstat,
4870         };
4871         struct rpc_message msg = {
4872                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4873                 .rpc_argp = &args,
4874                 .rpc_resp = &res,
4875         };
4876
4877         nfs_fattr_init(fsstat->fattr);
4878         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4879 }
4880
4881 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4882 {
4883         struct nfs4_exception exception = {
4884                 .interruptible = true,
4885         };
4886         int err;
4887         do {
4888                 err = nfs4_handle_exception(server,
4889                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4890                                 &exception);
4891         } while (exception.retry);
4892         return err;
4893 }
4894
4895 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4896                 struct nfs_fsinfo *fsinfo)
4897 {
4898         struct nfs4_fsinfo_arg args = {
4899                 .fh = fhandle,
4900                 .bitmask = server->attr_bitmask,
4901         };
4902         struct nfs4_fsinfo_res res = {
4903                 .fsinfo = fsinfo,
4904         };
4905         struct rpc_message msg = {
4906                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4907                 .rpc_argp = &args,
4908                 .rpc_resp = &res,
4909         };
4910
4911         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4912 }
4913
4914 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4915 {
4916         struct nfs4_exception exception = {
4917                 .interruptible = true,
4918         };
4919         unsigned long now = jiffies;
4920         int err;
4921
4922         do {
4923                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4924                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4925                 if (err == 0) {
4926                         nfs4_set_lease_period(server->nfs_client,
4927                                         fsinfo->lease_time * HZ,
4928                                         now);
4929                         break;
4930                 }
4931                 err = nfs4_handle_exception(server, err, &exception);
4932         } while (exception.retry);
4933         return err;
4934 }
4935
4936 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4937 {
4938         int error;
4939
4940         nfs_fattr_init(fsinfo->fattr);
4941         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4942         if (error == 0) {
4943                 /* block layout checks this! */
4944                 server->pnfs_blksize = fsinfo->blksize;
4945                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4946         }
4947
4948         return error;
4949 }
4950
4951 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4952                 struct nfs_pathconf *pathconf)
4953 {
4954         struct nfs4_pathconf_arg args = {
4955                 .fh = fhandle,
4956                 .bitmask = server->attr_bitmask,
4957         };
4958         struct nfs4_pathconf_res res = {
4959                 .pathconf = pathconf,
4960         };
4961         struct rpc_message msg = {
4962                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4963                 .rpc_argp = &args,
4964                 .rpc_resp = &res,
4965         };
4966
4967         /* None of the pathconf attributes are mandatory to implement */
4968         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4969                 memset(pathconf, 0, sizeof(*pathconf));
4970                 return 0;
4971         }
4972
4973         nfs_fattr_init(pathconf->fattr);
4974         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4975 }
4976
4977 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4978                 struct nfs_pathconf *pathconf)
4979 {
4980         struct nfs4_exception exception = {
4981                 .interruptible = true,
4982         };
4983         int err;
4984
4985         do {
4986                 err = nfs4_handle_exception(server,
4987                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4988                                 &exception);
4989         } while (exception.retry);
4990         return err;
4991 }
4992
4993 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4994                 const struct nfs_open_context *ctx,
4995                 const struct nfs_lock_context *l_ctx,
4996                 fmode_t fmode)
4997 {
4998         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4999 }
5000 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5001
5002 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5003                 const struct nfs_open_context *ctx,
5004                 const struct nfs_lock_context *l_ctx,
5005                 fmode_t fmode)
5006 {
5007         nfs4_stateid current_stateid;
5008
5009         /* If the current stateid represents a lost lock, then exit */
5010         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
5011                 return true;
5012         return nfs4_stateid_match(stateid, &current_stateid);
5013 }
5014
5015 static bool nfs4_error_stateid_expired(int err)
5016 {
5017         switch (err) {
5018         case -NFS4ERR_DELEG_REVOKED:
5019         case -NFS4ERR_ADMIN_REVOKED:
5020         case -NFS4ERR_BAD_STATEID:
5021         case -NFS4ERR_STALE_STATEID:
5022         case -NFS4ERR_OLD_STATEID:
5023         case -NFS4ERR_OPENMODE:
5024         case -NFS4ERR_EXPIRED:
5025                 return true;
5026         }
5027         return false;
5028 }
5029
5030 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5031 {
5032         struct nfs_server *server = NFS_SERVER(hdr->inode);
5033
5034         trace_nfs4_read(hdr, task->tk_status);
5035         if (task->tk_status < 0) {
5036                 struct nfs4_exception exception = {
5037                         .inode = hdr->inode,
5038                         .state = hdr->args.context->state,
5039                         .stateid = &hdr->args.stateid,
5040                 };
5041                 task->tk_status = nfs4_async_handle_exception(task,
5042                                 server, task->tk_status, &exception);
5043                 if (exception.retry) {
5044                         rpc_restart_call_prepare(task);
5045                         return -EAGAIN;
5046                 }
5047         }
5048
5049         if (task->tk_status > 0)
5050                 renew_lease(server, hdr->timestamp);
5051         return 0;
5052 }
5053
5054 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5055                 struct nfs_pgio_args *args)
5056 {
5057
5058         if (!nfs4_error_stateid_expired(task->tk_status) ||
5059                 nfs4_stateid_is_current(&args->stateid,
5060                                 args->context,
5061                                 args->lock_context,
5062                                 FMODE_READ))
5063                 return false;
5064         rpc_restart_call_prepare(task);
5065         return true;
5066 }
5067
5068 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5069 {
5070
5071         dprintk("--> %s\n", __func__);
5072
5073         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5074                 return -EAGAIN;
5075         if (nfs4_read_stateid_changed(task, &hdr->args))
5076                 return -EAGAIN;
5077         if (task->tk_status > 0)
5078                 nfs_invalidate_atime(hdr->inode);
5079         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5080                                     nfs4_read_done_cb(task, hdr);
5081 }
5082
5083 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5084                                  struct rpc_message *msg)
5085 {
5086         hdr->timestamp   = jiffies;
5087         if (!hdr->pgio_done_cb)
5088                 hdr->pgio_done_cb = nfs4_read_done_cb;
5089         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5090         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5091 }
5092
5093 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5094                                       struct nfs_pgio_header *hdr)
5095 {
5096         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5097                         &hdr->args.seq_args,
5098                         &hdr->res.seq_res,
5099                         task))
5100                 return 0;
5101         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5102                                 hdr->args.lock_context,
5103                                 hdr->rw_mode) == -EIO)
5104                 return -EIO;
5105         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5106                 return -EIO;
5107         return 0;
5108 }
5109
5110 static int nfs4_write_done_cb(struct rpc_task *task,
5111                               struct nfs_pgio_header *hdr)
5112 {
5113         struct inode *inode = hdr->inode;
5114
5115         trace_nfs4_write(hdr, task->tk_status);
5116         if (task->tk_status < 0) {
5117                 struct nfs4_exception exception = {
5118                         .inode = hdr->inode,
5119                         .state = hdr->args.context->state,
5120                         .stateid = &hdr->args.stateid,
5121                 };
5122                 task->tk_status = nfs4_async_handle_exception(task,
5123                                 NFS_SERVER(inode), task->tk_status,
5124                                 &exception);
5125                 if (exception.retry) {
5126                         rpc_restart_call_prepare(task);
5127                         return -EAGAIN;
5128                 }
5129         }
5130         if (task->tk_status >= 0) {
5131                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5132                 nfs_writeback_update_inode(hdr);
5133         }
5134         return 0;
5135 }
5136
5137 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5138                 struct nfs_pgio_args *args)
5139 {
5140
5141         if (!nfs4_error_stateid_expired(task->tk_status) ||
5142                 nfs4_stateid_is_current(&args->stateid,
5143                                 args->context,
5144                                 args->lock_context,
5145                                 FMODE_WRITE))
5146                 return false;
5147         rpc_restart_call_prepare(task);
5148         return true;
5149 }
5150
5151 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5152 {
5153         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5154                 return -EAGAIN;
5155         if (nfs4_write_stateid_changed(task, &hdr->args))
5156                 return -EAGAIN;
5157         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5158                 nfs4_write_done_cb(task, hdr);
5159 }
5160
5161 static
5162 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5163 {
5164         /* Don't request attributes for pNFS or O_DIRECT writes */
5165         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5166                 return false;
5167         /* Otherwise, request attributes if and only if we don't hold
5168          * a delegation
5169          */
5170         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5171 }
5172
5173 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5174                                   struct rpc_message *msg,
5175                                   struct rpc_clnt **clnt)
5176 {
5177         struct nfs_server *server = NFS_SERVER(hdr->inode);
5178
5179         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5180                 hdr->args.bitmask = NULL;
5181                 hdr->res.fattr = NULL;
5182         } else
5183                 hdr->args.bitmask = server->cache_consistency_bitmask;
5184
5185         if (!hdr->pgio_done_cb)
5186                 hdr->pgio_done_cb = nfs4_write_done_cb;
5187         hdr->res.server = server;
5188         hdr->timestamp   = jiffies;
5189
5190         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5191         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1, 0);
5192         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5193 }
5194
5195 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5196 {
5197         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5198                         &data->args.seq_args,
5199                         &data->res.seq_res,
5200                         task);
5201 }
5202
5203 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5204 {
5205         struct inode *inode = data->inode;
5206
5207         trace_nfs4_commit(data, task->tk_status);
5208         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5209                                     NULL, NULL) == -EAGAIN) {
5210                 rpc_restart_call_prepare(task);
5211                 return -EAGAIN;
5212         }
5213         return 0;
5214 }
5215
5216 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5217 {
5218         if (!nfs4_sequence_done(task, &data->res.seq_res))
5219                 return -EAGAIN;
5220         return data->commit_done_cb(task, data);
5221 }
5222
5223 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5224                                    struct rpc_clnt **clnt)
5225 {
5226         struct nfs_server *server = NFS_SERVER(data->inode);
5227
5228         if (data->commit_done_cb == NULL)
5229                 data->commit_done_cb = nfs4_commit_done_cb;
5230         data->res.server = server;
5231         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5232         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5233         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5234 }
5235
5236 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5237                                 struct nfs_commitres *res)
5238 {
5239         struct inode *dst_inode = file_inode(dst);
5240         struct nfs_server *server = NFS_SERVER(dst_inode);
5241         struct rpc_message msg = {
5242                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5243                 .rpc_argp = args,
5244                 .rpc_resp = res,
5245         };
5246
5247         args->fh = NFS_FH(dst_inode);
5248         return nfs4_call_sync(server->client, server, &msg,
5249                         &args->seq_args, &res->seq_res, 1);
5250 }
5251
5252 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5253 {
5254         struct nfs_commitargs args = {
5255                 .offset = offset,
5256                 .count = count,
5257         };
5258         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5259         struct nfs4_exception exception = { };
5260         int status;
5261
5262         do {
5263                 status = _nfs4_proc_commit(dst, &args, res);
5264                 status = nfs4_handle_exception(dst_server, status, &exception);
5265         } while (exception.retry);
5266
5267         return status;
5268 }
5269
5270 struct nfs4_renewdata {
5271         struct nfs_client       *client;
5272         unsigned long           timestamp;
5273 };
5274
5275 /*
5276  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5277  * standalone procedure for queueing an asynchronous RENEW.
5278  */
5279 static void nfs4_renew_release(void *calldata)
5280 {
5281         struct nfs4_renewdata *data = calldata;
5282         struct nfs_client *clp = data->client;
5283
5284         if (refcount_read(&clp->cl_count) > 1)
5285                 nfs4_schedule_state_renewal(clp);
5286         nfs_put_client(clp);
5287         kfree(data);
5288 }
5289
5290 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5291 {
5292         struct nfs4_renewdata *data = calldata;
5293         struct nfs_client *clp = data->client;
5294         unsigned long timestamp = data->timestamp;
5295
5296         trace_nfs4_renew_async(clp, task->tk_status);
5297         switch (task->tk_status) {
5298         case 0:
5299                 break;
5300         case -NFS4ERR_LEASE_MOVED:
5301                 nfs4_schedule_lease_moved_recovery(clp);
5302                 break;
5303         default:
5304                 /* Unless we're shutting down, schedule state recovery! */
5305                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5306                         return;
5307                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5308                         nfs4_schedule_lease_recovery(clp);
5309                         return;
5310                 }
5311                 nfs4_schedule_path_down_recovery(clp);
5312         }
5313         do_renew_lease(clp, timestamp);
5314 }
5315
5316 static const struct rpc_call_ops nfs4_renew_ops = {
5317         .rpc_call_done = nfs4_renew_done,
5318         .rpc_release = nfs4_renew_release,
5319 };
5320
5321 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5322 {
5323         struct rpc_message msg = {
5324                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5325                 .rpc_argp       = clp,
5326                 .rpc_cred       = cred,
5327         };
5328         struct nfs4_renewdata *data;
5329
5330         if (renew_flags == 0)
5331                 return 0;
5332         if (!refcount_inc_not_zero(&clp->cl_count))
5333                 return -EIO;
5334         data = kmalloc(sizeof(*data), GFP_NOFS);
5335         if (data == NULL) {
5336                 nfs_put_client(clp);
5337                 return -ENOMEM;
5338         }
5339         data->client = clp;
5340         data->timestamp = jiffies;
5341         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5342                         &nfs4_renew_ops, data);
5343 }
5344
5345 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5346 {
5347         struct rpc_message msg = {
5348                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5349                 .rpc_argp       = clp,
5350                 .rpc_cred       = cred,
5351         };
5352         unsigned long now = jiffies;
5353         int status;
5354
5355         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5356         if (status < 0)
5357                 return status;
5358         do_renew_lease(clp, now);
5359         return 0;
5360 }
5361
5362 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5363 {
5364         return server->caps & NFS_CAP_ACLS;
5365 }
5366
5367 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5368  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5369  * the stack.
5370  */
5371 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5372
5373 static int buf_to_pages_noslab(const void *buf, size_t buflen,
5374                 struct page **pages)
5375 {
5376         struct page *newpage, **spages;
5377         int rc = 0;
5378         size_t len;
5379         spages = pages;
5380
5381         do {
5382                 len = min_t(size_t, PAGE_SIZE, buflen);
5383                 newpage = alloc_page(GFP_KERNEL);
5384
5385                 if (newpage == NULL)
5386                         goto unwind;
5387                 memcpy(page_address(newpage), buf, len);
5388                 buf += len;
5389                 buflen -= len;
5390                 *pages++ = newpage;
5391                 rc++;
5392         } while (buflen != 0);
5393
5394         return rc;
5395
5396 unwind:
5397         for(; rc > 0; rc--)
5398                 __free_page(spages[rc-1]);
5399         return -ENOMEM;
5400 }
5401
5402 struct nfs4_cached_acl {
5403         int cached;
5404         size_t len;
5405         char data[0];
5406 };
5407
5408 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5409 {
5410         struct nfs_inode *nfsi = NFS_I(inode);
5411
5412         spin_lock(&inode->i_lock);
5413         kfree(nfsi->nfs4_acl);
5414         nfsi->nfs4_acl = acl;
5415         spin_unlock(&inode->i_lock);
5416 }
5417
5418 static void nfs4_zap_acl_attr(struct inode *inode)
5419 {
5420         nfs4_set_cached_acl(inode, NULL);
5421 }
5422
5423 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5424 {
5425         struct nfs_inode *nfsi = NFS_I(inode);
5426         struct nfs4_cached_acl *acl;
5427         int ret = -ENOENT;
5428
5429         spin_lock(&inode->i_lock);
5430         acl = nfsi->nfs4_acl;
5431         if (acl == NULL)
5432                 goto out;
5433         if (buf == NULL) /* user is just asking for length */
5434                 goto out_len;
5435         if (acl->cached == 0)
5436                 goto out;
5437         ret = -ERANGE; /* see getxattr(2) man page */
5438         if (acl->len > buflen)
5439                 goto out;
5440         memcpy(buf, acl->data, acl->len);
5441 out_len:
5442         ret = acl->len;
5443 out:
5444         spin_unlock(&inode->i_lock);
5445         return ret;
5446 }
5447
5448 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5449 {
5450         struct nfs4_cached_acl *acl;
5451         size_t buflen = sizeof(*acl) + acl_len;
5452
5453         if (buflen <= PAGE_SIZE) {
5454                 acl = kmalloc(buflen, GFP_KERNEL);
5455                 if (acl == NULL)
5456                         goto out;
5457                 acl->cached = 1;
5458                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5459         } else {
5460                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5461                 if (acl == NULL)
5462                         goto out;
5463                 acl->cached = 0;
5464         }
5465         acl->len = acl_len;
5466 out:
5467         nfs4_set_cached_acl(inode, acl);
5468 }
5469
5470 /*
5471  * The getxattr API returns the required buffer length when called with a
5472  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5473  * the required buf.  On a NULL buf, we send a page of data to the server
5474  * guessing that the ACL request can be serviced by a page. If so, we cache
5475  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5476  * the cache. If not so, we throw away the page, and cache the required
5477  * length. The next getxattr call will then produce another round trip to
5478  * the server, this time with the input buf of the required size.
5479  */
5480 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5481 {
5482         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
5483         struct nfs_getaclargs args = {
5484                 .fh = NFS_FH(inode),
5485                 .acl_pages = pages,
5486                 .acl_len = buflen,
5487         };
5488         struct nfs_getaclres res = {
5489                 .acl_len = buflen,
5490         };
5491         struct rpc_message msg = {
5492                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5493                 .rpc_argp = &args,
5494                 .rpc_resp = &res,
5495         };
5496         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5497         int ret = -ENOMEM, i;
5498
5499         if (npages > ARRAY_SIZE(pages))
5500                 return -ERANGE;
5501
5502         for (i = 0; i < npages; i++) {
5503                 pages[i] = alloc_page(GFP_KERNEL);
5504                 if (!pages[i])
5505                         goto out_free;
5506         }
5507
5508         /* for decoding across pages */
5509         res.acl_scratch = alloc_page(GFP_KERNEL);
5510         if (!res.acl_scratch)
5511                 goto out_free;
5512
5513         args.acl_len = npages * PAGE_SIZE;
5514
5515         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5516                 __func__, buf, buflen, npages, args.acl_len);
5517         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5518                              &msg, &args.seq_args, &res.seq_res, 0);
5519         if (ret)
5520                 goto out_free;
5521
5522         /* Handle the case where the passed-in buffer is too short */
5523         if (res.acl_flags & NFS4_ACL_TRUNC) {
5524                 /* Did the user only issue a request for the acl length? */
5525                 if (buf == NULL)
5526                         goto out_ok;
5527                 ret = -ERANGE;
5528                 goto out_free;
5529         }
5530         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5531         if (buf) {
5532                 if (res.acl_len > buflen) {
5533                         ret = -ERANGE;
5534                         goto out_free;
5535                 }
5536                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5537         }
5538 out_ok:
5539         ret = res.acl_len;
5540 out_free:
5541         for (i = 0; i < npages; i++)
5542                 if (pages[i])
5543                         __free_page(pages[i]);
5544         if (res.acl_scratch)
5545                 __free_page(res.acl_scratch);
5546         return ret;
5547 }
5548
5549 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5550 {
5551         struct nfs4_exception exception = {
5552                 .interruptible = true,
5553         };
5554         ssize_t ret;
5555         do {
5556                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5557                 trace_nfs4_get_acl(inode, ret);
5558                 if (ret >= 0)
5559                         break;
5560                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5561         } while (exception.retry);
5562         return ret;
5563 }
5564
5565 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5566 {
5567         struct nfs_server *server = NFS_SERVER(inode);
5568         int ret;
5569
5570         if (!nfs4_server_supports_acls(server))
5571                 return -EOPNOTSUPP;
5572         ret = nfs_revalidate_inode(server, inode);
5573         if (ret < 0)
5574                 return ret;
5575         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5576                 nfs_zap_acl_cache(inode);
5577         ret = nfs4_read_cached_acl(inode, buf, buflen);
5578         if (ret != -ENOENT)
5579                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5580                  * but no cached acl data, just the acl length */
5581                 return ret;
5582         return nfs4_get_acl_uncached(inode, buf, buflen);
5583 }
5584
5585 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5586 {
5587         struct nfs_server *server = NFS_SERVER(inode);
5588         struct page *pages[NFS4ACL_MAXPAGES];
5589         struct nfs_setaclargs arg = {
5590                 .fh             = NFS_FH(inode),
5591                 .acl_pages      = pages,
5592                 .acl_len        = buflen,
5593         };
5594         struct nfs_setaclres res;
5595         struct rpc_message msg = {
5596                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5597                 .rpc_argp       = &arg,
5598                 .rpc_resp       = &res,
5599         };
5600         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5601         int ret, i;
5602
5603         if (!nfs4_server_supports_acls(server))
5604                 return -EOPNOTSUPP;
5605         if (npages > ARRAY_SIZE(pages))
5606                 return -ERANGE;
5607         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5608         if (i < 0)
5609                 return i;
5610         nfs4_inode_make_writeable(inode);
5611         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5612
5613         /*
5614          * Free each page after tx, so the only ref left is
5615          * held by the network stack
5616          */
5617         for (; i > 0; i--)
5618                 put_page(pages[i-1]);
5619
5620         /*
5621          * Acl update can result in inode attribute update.
5622          * so mark the attribute cache invalid.
5623          */
5624         spin_lock(&inode->i_lock);
5625         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5626                 | NFS_INO_INVALID_CTIME
5627                 | NFS_INO_REVAL_FORCED;
5628         spin_unlock(&inode->i_lock);
5629         nfs_access_zap_cache(inode);
5630         nfs_zap_acl_cache(inode);
5631         return ret;
5632 }
5633
5634 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5635 {
5636         struct nfs4_exception exception = { };
5637         int err;
5638         do {
5639                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5640                 trace_nfs4_set_acl(inode, err);
5641                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5642                                 &exception);
5643         } while (exception.retry);
5644         return err;
5645 }
5646
5647 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5648 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5649                                         size_t buflen)
5650 {
5651         struct nfs_server *server = NFS_SERVER(inode);
5652         struct nfs_fattr fattr;
5653         struct nfs4_label label = {0, 0, buflen, buf};
5654
5655         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5656         struct nfs4_getattr_arg arg = {
5657                 .fh             = NFS_FH(inode),
5658                 .bitmask        = bitmask,
5659         };
5660         struct nfs4_getattr_res res = {
5661                 .fattr          = &fattr,
5662                 .label          = &label,
5663                 .server         = server,
5664         };
5665         struct rpc_message msg = {
5666                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5667                 .rpc_argp       = &arg,
5668                 .rpc_resp       = &res,
5669         };
5670         int ret;
5671
5672         nfs_fattr_init(&fattr);
5673
5674         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5675         if (ret)
5676                 return ret;
5677         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5678                 return -ENOENT;
5679         if (buflen < label.len)
5680                 return -ERANGE;
5681         return 0;
5682 }
5683
5684 static int nfs4_get_security_label(struct inode *inode, void *buf,
5685                                         size_t buflen)
5686 {
5687         struct nfs4_exception exception = {
5688                 .interruptible = true,
5689         };
5690         int err;
5691
5692         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5693                 return -EOPNOTSUPP;
5694
5695         do {
5696                 err = _nfs4_get_security_label(inode, buf, buflen);
5697                 trace_nfs4_get_security_label(inode, err);
5698                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5699                                 &exception);
5700         } while (exception.retry);
5701         return err;
5702 }
5703
5704 static int _nfs4_do_set_security_label(struct inode *inode,
5705                 struct nfs4_label *ilabel,
5706                 struct nfs_fattr *fattr,
5707                 struct nfs4_label *olabel)
5708 {
5709
5710         struct iattr sattr = {0};
5711         struct nfs_server *server = NFS_SERVER(inode);
5712         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5713         struct nfs_setattrargs arg = {
5714                 .fh             = NFS_FH(inode),
5715                 .iap            = &sattr,
5716                 .server         = server,
5717                 .bitmask        = bitmask,
5718                 .label          = ilabel,
5719         };
5720         struct nfs_setattrres res = {
5721                 .fattr          = fattr,
5722                 .label          = olabel,
5723                 .server         = server,
5724         };
5725         struct rpc_message msg = {
5726                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5727                 .rpc_argp       = &arg,
5728                 .rpc_resp       = &res,
5729         };
5730         int status;
5731
5732         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5733
5734         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5735         if (status)
5736                 dprintk("%s failed: %d\n", __func__, status);
5737
5738         return status;
5739 }
5740
5741 static int nfs4_do_set_security_label(struct inode *inode,
5742                 struct nfs4_label *ilabel,
5743                 struct nfs_fattr *fattr,
5744                 struct nfs4_label *olabel)
5745 {
5746         struct nfs4_exception exception = { };
5747         int err;
5748
5749         do {
5750                 err = _nfs4_do_set_security_label(inode, ilabel,
5751                                 fattr, olabel);
5752                 trace_nfs4_set_security_label(inode, err);
5753                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5754                                 &exception);
5755         } while (exception.retry);
5756         return err;
5757 }
5758
5759 static int
5760 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5761 {
5762         struct nfs4_label ilabel, *olabel = NULL;
5763         struct nfs_fattr fattr;
5764         int status;
5765
5766         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5767                 return -EOPNOTSUPP;
5768
5769         nfs_fattr_init(&fattr);
5770
5771         ilabel.pi = 0;
5772         ilabel.lfs = 0;
5773         ilabel.label = (char *)buf;
5774         ilabel.len = buflen;
5775
5776         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5777         if (IS_ERR(olabel)) {
5778                 status = -PTR_ERR(olabel);
5779                 goto out;
5780         }
5781
5782         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5783         if (status == 0)
5784                 nfs_setsecurity(inode, &fattr, olabel);
5785
5786         nfs4_label_free(olabel);
5787 out:
5788         return status;
5789 }
5790 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5791
5792
5793 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5794                                     nfs4_verifier *bootverf)
5795 {
5796         __be32 verf[2];
5797
5798         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5799                 /* An impossible timestamp guarantees this value
5800                  * will never match a generated boot time. */
5801                 verf[0] = cpu_to_be32(U32_MAX);
5802                 verf[1] = cpu_to_be32(U32_MAX);
5803         } else {
5804                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5805                 u64 ns = ktime_to_ns(nn->boot_time);
5806
5807                 verf[0] = cpu_to_be32(ns >> 32);
5808                 verf[1] = cpu_to_be32(ns);
5809         }
5810         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5811 }
5812
5813 static int
5814 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5815 {
5816         size_t len;
5817         char *str;
5818
5819         if (clp->cl_owner_id != NULL)
5820                 return 0;
5821
5822         rcu_read_lock();
5823         len = 14 +
5824                 strlen(clp->cl_rpcclient->cl_nodename) +
5825                 1 +
5826                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5827                 1;
5828         rcu_read_unlock();
5829         if (nfs4_client_id_uniquifier[0] != '\0')
5830                 len += strlen(nfs4_client_id_uniquifier) + 1;
5831         if (len > NFS4_OPAQUE_LIMIT + 1)
5832                 return -EINVAL;
5833
5834         /*
5835          * Since this string is allocated at mount time, and held until the
5836          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5837          * about a memory-reclaim deadlock.
5838          */
5839         str = kmalloc(len, GFP_KERNEL);
5840         if (!str)
5841                 return -ENOMEM;
5842
5843         rcu_read_lock();
5844         if (nfs4_client_id_uniquifier[0] != '\0')
5845                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
5846                           clp->cl_rpcclient->cl_nodename,
5847                           nfs4_client_id_uniquifier,
5848                           rpc_peeraddr2str(clp->cl_rpcclient,
5849                                            RPC_DISPLAY_ADDR));
5850         else
5851                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
5852                           clp->cl_rpcclient->cl_nodename,
5853                           rpc_peeraddr2str(clp->cl_rpcclient,
5854                                            RPC_DISPLAY_ADDR));
5855         rcu_read_unlock();
5856
5857         clp->cl_owner_id = str;
5858         return 0;
5859 }
5860
5861 static int
5862 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5863 {
5864         size_t len;
5865         char *str;
5866
5867         len = 10 + 10 + 1 + 10 + 1 +
5868                 strlen(nfs4_client_id_uniquifier) + 1 +
5869                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5870
5871         if (len > NFS4_OPAQUE_LIMIT + 1)
5872                 return -EINVAL;
5873
5874         /*
5875          * Since this string is allocated at mount time, and held until the
5876          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5877          * about a memory-reclaim deadlock.
5878          */
5879         str = kmalloc(len, GFP_KERNEL);
5880         if (!str)
5881                 return -ENOMEM;
5882
5883         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5884                         clp->rpc_ops->version, clp->cl_minorversion,
5885                         nfs4_client_id_uniquifier,
5886                         clp->cl_rpcclient->cl_nodename);
5887         clp->cl_owner_id = str;
5888         return 0;
5889 }
5890
5891 static int
5892 nfs4_init_uniform_client_string(struct nfs_client *clp)
5893 {
5894         size_t len;
5895         char *str;
5896
5897         if (clp->cl_owner_id != NULL)
5898                 return 0;
5899
5900         if (nfs4_client_id_uniquifier[0] != '\0')
5901                 return nfs4_init_uniquifier_client_string(clp);
5902
5903         len = 10 + 10 + 1 + 10 + 1 +
5904                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5905
5906         if (len > NFS4_OPAQUE_LIMIT + 1)
5907                 return -EINVAL;
5908
5909         /*
5910          * Since this string is allocated at mount time, and held until the
5911          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5912          * about a memory-reclaim deadlock.
5913          */
5914         str = kmalloc(len, GFP_KERNEL);
5915         if (!str)
5916                 return -ENOMEM;
5917
5918         scnprintf(str, len, "Linux NFSv%u.%u %s",
5919                         clp->rpc_ops->version, clp->cl_minorversion,
5920                         clp->cl_rpcclient->cl_nodename);
5921         clp->cl_owner_id = str;
5922         return 0;
5923 }
5924
5925 /*
5926  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5927  * services.  Advertise one based on the address family of the
5928  * clientaddr.
5929  */
5930 static unsigned int
5931 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5932 {
5933         if (strchr(clp->cl_ipaddr, ':') != NULL)
5934                 return scnprintf(buf, len, "tcp6");
5935         else
5936                 return scnprintf(buf, len, "tcp");
5937 }
5938
5939 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5940 {
5941         struct nfs4_setclientid *sc = calldata;
5942
5943         if (task->tk_status == 0)
5944                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5945 }
5946
5947 static const struct rpc_call_ops nfs4_setclientid_ops = {
5948         .rpc_call_done = nfs4_setclientid_done,
5949 };
5950
5951 /**
5952  * nfs4_proc_setclientid - Negotiate client ID
5953  * @clp: state data structure
5954  * @program: RPC program for NFSv4 callback service
5955  * @port: IP port number for NFS4 callback service
5956  * @cred: credential to use for this call
5957  * @res: where to place the result
5958  *
5959  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5960  */
5961 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5962                 unsigned short port, const struct cred *cred,
5963                 struct nfs4_setclientid_res *res)
5964 {
5965         nfs4_verifier sc_verifier;
5966         struct nfs4_setclientid setclientid = {
5967                 .sc_verifier = &sc_verifier,
5968                 .sc_prog = program,
5969                 .sc_clnt = clp,
5970         };
5971         struct rpc_message msg = {
5972                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5973                 .rpc_argp = &setclientid,
5974                 .rpc_resp = res,
5975                 .rpc_cred = cred,
5976         };
5977         struct rpc_task *task;
5978         struct rpc_task_setup task_setup_data = {
5979                 .rpc_client = clp->cl_rpcclient,
5980                 .rpc_message = &msg,
5981                 .callback_ops = &nfs4_setclientid_ops,
5982                 .callback_data = &setclientid,
5983                 .flags = RPC_TASK_TIMEOUT,
5984         };
5985         int status;
5986
5987         /* nfs_client_id4 */
5988         nfs4_init_boot_verifier(clp, &sc_verifier);
5989
5990         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5991                 status = nfs4_init_uniform_client_string(clp);
5992         else
5993                 status = nfs4_init_nonuniform_client_string(clp);
5994
5995         if (status)
5996                 goto out;
5997
5998         /* cb_client4 */
5999         setclientid.sc_netid_len =
6000                                 nfs4_init_callback_netid(clp,
6001                                                 setclientid.sc_netid,
6002                                                 sizeof(setclientid.sc_netid));
6003         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6004                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6005                                 clp->cl_ipaddr, port >> 8, port & 255);
6006
6007         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6008                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6009                 clp->cl_owner_id);
6010         task = rpc_run_task(&task_setup_data);
6011         if (IS_ERR(task)) {
6012                 status = PTR_ERR(task);
6013                 goto out;
6014         }
6015         status = task->tk_status;
6016         if (setclientid.sc_cred) {
6017                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6018                 put_rpccred(setclientid.sc_cred);
6019         }
6020         rpc_put_task(task);
6021 out:
6022         trace_nfs4_setclientid(clp, status);
6023         dprintk("NFS reply setclientid: %d\n", status);
6024         return status;
6025 }
6026
6027 /**
6028  * nfs4_proc_setclientid_confirm - Confirm client ID
6029  * @clp: state data structure
6030  * @arg: result of a previous SETCLIENTID
6031  * @cred: credential to use for this call
6032  *
6033  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6034  */
6035 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6036                 struct nfs4_setclientid_res *arg,
6037                 const struct cred *cred)
6038 {
6039         struct rpc_message msg = {
6040                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6041                 .rpc_argp = arg,
6042                 .rpc_cred = cred,
6043         };
6044         int status;
6045
6046         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6047                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6048                 clp->cl_clientid);
6049         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6050         trace_nfs4_setclientid_confirm(clp, status);
6051         dprintk("NFS reply setclientid_confirm: %d\n", status);
6052         return status;
6053 }
6054
6055 struct nfs4_delegreturndata {
6056         struct nfs4_delegreturnargs args;
6057         struct nfs4_delegreturnres res;
6058         struct nfs_fh fh;
6059         nfs4_stateid stateid;
6060         unsigned long timestamp;
6061         struct {
6062                 struct nfs4_layoutreturn_args arg;
6063                 struct nfs4_layoutreturn_res res;
6064                 struct nfs4_xdr_opaque_data ld_private;
6065                 u32 roc_barrier;
6066                 bool roc;
6067         } lr;
6068         struct nfs_fattr fattr;
6069         int rpc_status;
6070         struct inode *inode;
6071 };
6072
6073 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6074 {
6075         struct nfs4_delegreturndata *data = calldata;
6076         struct nfs4_exception exception = {
6077                 .inode = data->inode,
6078                 .stateid = &data->stateid,
6079         };
6080
6081         if (!nfs4_sequence_done(task, &data->res.seq_res))
6082                 return;
6083
6084         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6085
6086         /* Handle Layoutreturn errors */
6087         if (data->args.lr_args && task->tk_status != 0) {
6088                 switch(data->res.lr_ret) {
6089                 default:
6090                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6091                         break;
6092                 case 0:
6093                         data->args.lr_args = NULL;
6094                         data->res.lr_res = NULL;
6095                         break;
6096                 case -NFS4ERR_OLD_STATEID:
6097                         if (nfs4_layoutreturn_refresh_stateid(&data->args.lr_args->stateid,
6098                                                 &data->args.lr_args->range,
6099                                                 data->inode))
6100                                 goto lr_restart;
6101                         /* Fallthrough */
6102                 case -NFS4ERR_ADMIN_REVOKED:
6103                 case -NFS4ERR_DELEG_REVOKED:
6104                 case -NFS4ERR_EXPIRED:
6105                 case -NFS4ERR_BAD_STATEID:
6106                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
6107                 case -NFS4ERR_WRONG_CRED:
6108                         data->args.lr_args = NULL;
6109                         data->res.lr_res = NULL;
6110                         goto lr_restart;
6111                 }
6112         }
6113
6114         switch (task->tk_status) {
6115         case 0:
6116                 renew_lease(data->res.server, data->timestamp);
6117                 break;
6118         case -NFS4ERR_ADMIN_REVOKED:
6119         case -NFS4ERR_DELEG_REVOKED:
6120         case -NFS4ERR_EXPIRED:
6121                 nfs4_free_revoked_stateid(data->res.server,
6122                                 data->args.stateid,
6123                                 task->tk_msg.rpc_cred);
6124                 /* Fallthrough */
6125         case -NFS4ERR_BAD_STATEID:
6126         case -NFS4ERR_STALE_STATEID:
6127                 task->tk_status = 0;
6128                 break;
6129         case -NFS4ERR_OLD_STATEID:
6130                 if (nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6131                         goto out_restart;
6132                 task->tk_status = 0;
6133                 break;
6134         case -NFS4ERR_ACCESS:
6135                 if (data->args.bitmask) {
6136                         data->args.bitmask = NULL;
6137                         data->res.fattr = NULL;
6138                         goto out_restart;
6139                 }
6140                 /* Fallthrough */
6141         default:
6142                 task->tk_status = nfs4_async_handle_exception(task,
6143                                 data->res.server, task->tk_status,
6144                                 &exception);
6145                 if (exception.retry)
6146                         goto out_restart;
6147         }
6148         data->rpc_status = task->tk_status;
6149         return;
6150 lr_restart:
6151         data->res.lr_ret = 0;
6152 out_restart:
6153         task->tk_status = 0;
6154         rpc_restart_call_prepare(task);
6155 }
6156
6157 static void nfs4_delegreturn_release(void *calldata)
6158 {
6159         struct nfs4_delegreturndata *data = calldata;
6160         struct inode *inode = data->inode;
6161
6162         if (inode) {
6163                 if (data->lr.roc)
6164                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
6165                                         data->res.lr_ret);
6166                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6167                 nfs_iput_and_deactive(inode);
6168         }
6169         kfree(calldata);
6170 }
6171
6172 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6173 {
6174         struct nfs4_delegreturndata *d_data;
6175         struct pnfs_layout_hdr *lo;
6176
6177         d_data = (struct nfs4_delegreturndata *)data;
6178
6179         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
6180                 return;
6181
6182         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6183         if (lo && !pnfs_layout_is_valid(lo)) {
6184                 d_data->args.lr_args = NULL;
6185                 d_data->res.lr_res = NULL;
6186         }
6187
6188         nfs4_setup_sequence(d_data->res.server->nfs_client,
6189                         &d_data->args.seq_args,
6190                         &d_data->res.seq_res,
6191                         task);
6192 }
6193
6194 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6195         .rpc_call_prepare = nfs4_delegreturn_prepare,
6196         .rpc_call_done = nfs4_delegreturn_done,
6197         .rpc_release = nfs4_delegreturn_release,
6198 };
6199
6200 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6201 {
6202         struct nfs4_delegreturndata *data;
6203         struct nfs_server *server = NFS_SERVER(inode);
6204         struct rpc_task *task;
6205         struct rpc_message msg = {
6206                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6207                 .rpc_cred = cred,
6208         };
6209         struct rpc_task_setup task_setup_data = {
6210                 .rpc_client = server->client,
6211                 .rpc_message = &msg,
6212                 .callback_ops = &nfs4_delegreturn_ops,
6213                 .flags = RPC_TASK_ASYNC,
6214         };
6215         int status = 0;
6216
6217         data = kzalloc(sizeof(*data), GFP_NOFS);
6218         if (data == NULL)
6219                 return -ENOMEM;
6220         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6221
6222         nfs4_state_protect(server->nfs_client,
6223                         NFS_SP4_MACH_CRED_CLEANUP,
6224                         &task_setup_data.rpc_client, &msg);
6225
6226         data->args.fhandle = &data->fh;
6227         data->args.stateid = &data->stateid;
6228         data->args.bitmask = server->cache_consistency_bitmask;
6229         nfs_copy_fh(&data->fh, NFS_FH(inode));
6230         nfs4_stateid_copy(&data->stateid, stateid);
6231         data->res.fattr = &data->fattr;
6232         data->res.server = server;
6233         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6234         data->lr.arg.ld_private = &data->lr.ld_private;
6235         nfs_fattr_init(data->res.fattr);
6236         data->timestamp = jiffies;
6237         data->rpc_status = 0;
6238         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
6239         data->inode = nfs_igrab_and_active(inode);
6240         if (data->inode) {
6241                 if (data->lr.roc) {
6242                         data->args.lr_args = &data->lr.arg;
6243                         data->res.lr_res = &data->lr.res;
6244                 }
6245         } else if (data->lr.roc) {
6246                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
6247                 data->lr.roc = false;
6248         }
6249
6250         task_setup_data.callback_data = data;
6251         msg.rpc_argp = &data->args;
6252         msg.rpc_resp = &data->res;
6253         task = rpc_run_task(&task_setup_data);
6254         if (IS_ERR(task))
6255                 return PTR_ERR(task);
6256         if (!issync)
6257                 goto out;
6258         status = rpc_wait_for_completion_task(task);
6259         if (status != 0)
6260                 goto out;
6261         status = data->rpc_status;
6262 out:
6263         rpc_put_task(task);
6264         return status;
6265 }
6266
6267 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6268 {
6269         struct nfs_server *server = NFS_SERVER(inode);
6270         struct nfs4_exception exception = { };
6271         int err;
6272         do {
6273                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6274                 trace_nfs4_delegreturn(inode, stateid, err);
6275                 switch (err) {
6276                         case -NFS4ERR_STALE_STATEID:
6277                         case -NFS4ERR_EXPIRED:
6278                         case 0:
6279                                 return 0;
6280                 }
6281                 err = nfs4_handle_exception(server, err, &exception);
6282         } while (exception.retry);
6283         return err;
6284 }
6285
6286 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6287 {
6288         struct inode *inode = state->inode;
6289         struct nfs_server *server = NFS_SERVER(inode);
6290         struct nfs_client *clp = server->nfs_client;
6291         struct nfs_lockt_args arg = {
6292                 .fh = NFS_FH(inode),
6293                 .fl = request,
6294         };
6295         struct nfs_lockt_res res = {
6296                 .denied = request,
6297         };
6298         struct rpc_message msg = {
6299                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6300                 .rpc_argp       = &arg,
6301                 .rpc_resp       = &res,
6302                 .rpc_cred       = state->owner->so_cred,
6303         };
6304         struct nfs4_lock_state *lsp;
6305         int status;
6306
6307         arg.lock_owner.clientid = clp->cl_clientid;
6308         status = nfs4_set_lock_state(state, request);
6309         if (status != 0)
6310                 goto out;
6311         lsp = request->fl_u.nfs4_fl.owner;
6312         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6313         arg.lock_owner.s_dev = server->s_dev;
6314         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6315         switch (status) {
6316                 case 0:
6317                         request->fl_type = F_UNLCK;
6318                         break;
6319                 case -NFS4ERR_DENIED:
6320                         status = 0;
6321         }
6322         request->fl_ops->fl_release_private(request);
6323         request->fl_ops = NULL;
6324 out:
6325         return status;
6326 }
6327
6328 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6329 {
6330         struct nfs4_exception exception = {
6331                 .interruptible = true,
6332         };
6333         int err;
6334
6335         do {
6336                 err = _nfs4_proc_getlk(state, cmd, request);
6337                 trace_nfs4_get_lock(request, state, cmd, err);
6338                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6339                                 &exception);
6340         } while (exception.retry);
6341         return err;
6342 }
6343
6344 struct nfs4_unlockdata {
6345         struct nfs_locku_args arg;
6346         struct nfs_locku_res res;
6347         struct nfs4_lock_state *lsp;
6348         struct nfs_open_context *ctx;
6349         struct nfs_lock_context *l_ctx;
6350         struct file_lock fl;
6351         struct nfs_server *server;
6352         unsigned long timestamp;
6353 };
6354
6355 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6356                 struct nfs_open_context *ctx,
6357                 struct nfs4_lock_state *lsp,
6358                 struct nfs_seqid *seqid)
6359 {
6360         struct nfs4_unlockdata *p;
6361         struct inode *inode = lsp->ls_state->inode;
6362
6363         p = kzalloc(sizeof(*p), GFP_NOFS);
6364         if (p == NULL)
6365                 return NULL;
6366         p->arg.fh = NFS_FH(inode);
6367         p->arg.fl = &p->fl;
6368         p->arg.seqid = seqid;
6369         p->res.seqid = seqid;
6370         p->lsp = lsp;
6371         /* Ensure we don't close file until we're done freeing locks! */
6372         p->ctx = get_nfs_open_context(ctx);
6373         p->l_ctx = nfs_get_lock_context(ctx);
6374         locks_init_lock(&p->fl);
6375         locks_copy_lock(&p->fl, fl);
6376         p->server = NFS_SERVER(inode);
6377         return p;
6378 }
6379
6380 static void nfs4_locku_release_calldata(void *data)
6381 {
6382         struct nfs4_unlockdata *calldata = data;
6383         nfs_free_seqid(calldata->arg.seqid);
6384         nfs4_put_lock_state(calldata->lsp);
6385         nfs_put_lock_context(calldata->l_ctx);
6386         put_nfs_open_context(calldata->ctx);
6387         kfree(calldata);
6388 }
6389
6390 static void nfs4_locku_done(struct rpc_task *task, void *data)
6391 {
6392         struct nfs4_unlockdata *calldata = data;
6393         struct nfs4_exception exception = {
6394                 .inode = calldata->lsp->ls_state->inode,
6395                 .stateid = &calldata->arg.stateid,
6396         };
6397
6398         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6399                 return;
6400         switch (task->tk_status) {
6401                 case 0:
6402                         renew_lease(calldata->server, calldata->timestamp);
6403                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6404                         if (nfs4_update_lock_stateid(calldata->lsp,
6405                                         &calldata->res.stateid))
6406                                 break;
6407                         /* Fall through */
6408                 case -NFS4ERR_ADMIN_REVOKED:
6409                 case -NFS4ERR_EXPIRED:
6410                         nfs4_free_revoked_stateid(calldata->server,
6411                                         &calldata->arg.stateid,
6412                                         task->tk_msg.rpc_cred);
6413                         /* Fall through */
6414                 case -NFS4ERR_BAD_STATEID:
6415                 case -NFS4ERR_OLD_STATEID:
6416                 case -NFS4ERR_STALE_STATEID:
6417                         if (!nfs4_stateid_match(&calldata->arg.stateid,
6418                                                 &calldata->lsp->ls_stateid))
6419                                 rpc_restart_call_prepare(task);
6420                         break;
6421                 default:
6422                         task->tk_status = nfs4_async_handle_exception(task,
6423                                         calldata->server, task->tk_status,
6424                                         &exception);
6425                         if (exception.retry)
6426                                 rpc_restart_call_prepare(task);
6427         }
6428         nfs_release_seqid(calldata->arg.seqid);
6429 }
6430
6431 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6432 {
6433         struct nfs4_unlockdata *calldata = data;
6434
6435         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6436                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6437                 return;
6438
6439         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6440                 goto out_wait;
6441         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
6442         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6443                 /* Note: exit _without_ running nfs4_locku_done */
6444                 goto out_no_action;
6445         }
6446         calldata->timestamp = jiffies;
6447         if (nfs4_setup_sequence(calldata->server->nfs_client,
6448                                 &calldata->arg.seq_args,
6449                                 &calldata->res.seq_res,
6450                                 task) != 0)
6451                 nfs_release_seqid(calldata->arg.seqid);
6452         return;
6453 out_no_action:
6454         task->tk_action = NULL;
6455 out_wait:
6456         nfs4_sequence_done(task, &calldata->res.seq_res);
6457 }
6458
6459 static const struct rpc_call_ops nfs4_locku_ops = {
6460         .rpc_call_prepare = nfs4_locku_prepare,
6461         .rpc_call_done = nfs4_locku_done,
6462         .rpc_release = nfs4_locku_release_calldata,
6463 };
6464
6465 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6466                 struct nfs_open_context *ctx,
6467                 struct nfs4_lock_state *lsp,
6468                 struct nfs_seqid *seqid)
6469 {
6470         struct nfs4_unlockdata *data;
6471         struct rpc_message msg = {
6472                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6473                 .rpc_cred = ctx->cred,
6474         };
6475         struct rpc_task_setup task_setup_data = {
6476                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6477                 .rpc_message = &msg,
6478                 .callback_ops = &nfs4_locku_ops,
6479                 .workqueue = nfsiod_workqueue,
6480                 .flags = RPC_TASK_ASYNC,
6481         };
6482
6483         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6484                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6485
6486         /* Ensure this is an unlock - when canceling a lock, the
6487          * canceled lock is passed in, and it won't be an unlock.
6488          */
6489         fl->fl_type = F_UNLCK;
6490         if (fl->fl_flags & FL_CLOSE)
6491                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6492
6493         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6494         if (data == NULL) {
6495                 nfs_free_seqid(seqid);
6496                 return ERR_PTR(-ENOMEM);
6497         }
6498
6499         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6500         msg.rpc_argp = &data->arg;
6501         msg.rpc_resp = &data->res;
6502         task_setup_data.callback_data = data;
6503         return rpc_run_task(&task_setup_data);
6504 }
6505
6506 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6507 {
6508         struct inode *inode = state->inode;
6509         struct nfs4_state_owner *sp = state->owner;
6510         struct nfs_inode *nfsi = NFS_I(inode);
6511         struct nfs_seqid *seqid;
6512         struct nfs4_lock_state *lsp;
6513         struct rpc_task *task;
6514         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6515         int status = 0;
6516         unsigned char fl_flags = request->fl_flags;
6517
6518         status = nfs4_set_lock_state(state, request);
6519         /* Unlock _before_ we do the RPC call */
6520         request->fl_flags |= FL_EXISTS;
6521         /* Exclude nfs_delegation_claim_locks() */
6522         mutex_lock(&sp->so_delegreturn_mutex);
6523         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6524         down_read(&nfsi->rwsem);
6525         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6526                 up_read(&nfsi->rwsem);
6527                 mutex_unlock(&sp->so_delegreturn_mutex);
6528                 goto out;
6529         }
6530         up_read(&nfsi->rwsem);
6531         mutex_unlock(&sp->so_delegreturn_mutex);
6532         if (status != 0)
6533                 goto out;
6534         /* Is this a delegated lock? */
6535         lsp = request->fl_u.nfs4_fl.owner;
6536         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6537                 goto out;
6538         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6539         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6540         status = -ENOMEM;
6541         if (IS_ERR(seqid))
6542                 goto out;
6543         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6544         status = PTR_ERR(task);
6545         if (IS_ERR(task))
6546                 goto out;
6547         status = rpc_wait_for_completion_task(task);
6548         rpc_put_task(task);
6549 out:
6550         request->fl_flags = fl_flags;
6551         trace_nfs4_unlock(request, state, F_SETLK, status);
6552         return status;
6553 }
6554
6555 struct nfs4_lockdata {
6556         struct nfs_lock_args arg;
6557         struct nfs_lock_res res;
6558         struct nfs4_lock_state *lsp;
6559         struct nfs_open_context *ctx;
6560         struct file_lock fl;
6561         unsigned long timestamp;
6562         int rpc_status;
6563         int cancelled;
6564         struct nfs_server *server;
6565 };
6566
6567 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6568                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6569                 gfp_t gfp_mask)
6570 {
6571         struct nfs4_lockdata *p;
6572         struct inode *inode = lsp->ls_state->inode;
6573         struct nfs_server *server = NFS_SERVER(inode);
6574         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6575
6576         p = kzalloc(sizeof(*p), gfp_mask);
6577         if (p == NULL)
6578                 return NULL;
6579
6580         p->arg.fh = NFS_FH(inode);
6581         p->arg.fl = &p->fl;
6582         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6583         if (IS_ERR(p->arg.open_seqid))
6584                 goto out_free;
6585         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6586         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6587         if (IS_ERR(p->arg.lock_seqid))
6588                 goto out_free_seqid;
6589         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6590         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6591         p->arg.lock_owner.s_dev = server->s_dev;
6592         p->res.lock_seqid = p->arg.lock_seqid;
6593         p->lsp = lsp;
6594         p->server = server;
6595         p->ctx = get_nfs_open_context(ctx);
6596         locks_init_lock(&p->fl);
6597         locks_copy_lock(&p->fl, fl);
6598         return p;
6599 out_free_seqid:
6600         nfs_free_seqid(p->arg.open_seqid);
6601 out_free:
6602         kfree(p);
6603         return NULL;
6604 }
6605
6606 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6607 {
6608         struct nfs4_lockdata *data = calldata;
6609         struct nfs4_state *state = data->lsp->ls_state;
6610
6611         dprintk("%s: begin!\n", __func__);
6612         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6613                 goto out_wait;
6614         /* Do we need to do an open_to_lock_owner? */
6615         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6616                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6617                         goto out_release_lock_seqid;
6618                 }
6619                 nfs4_stateid_copy(&data->arg.open_stateid,
6620                                 &state->open_stateid);
6621                 data->arg.new_lock_owner = 1;
6622                 data->res.open_seqid = data->arg.open_seqid;
6623         } else {
6624                 data->arg.new_lock_owner = 0;
6625                 nfs4_stateid_copy(&data->arg.lock_stateid,
6626                                 &data->lsp->ls_stateid);
6627         }
6628         if (!nfs4_valid_open_stateid(state)) {
6629                 data->rpc_status = -EBADF;
6630                 task->tk_action = NULL;
6631                 goto out_release_open_seqid;
6632         }
6633         data->timestamp = jiffies;
6634         if (nfs4_setup_sequence(data->server->nfs_client,
6635                                 &data->arg.seq_args,
6636                                 &data->res.seq_res,
6637                                 task) == 0)
6638                 return;
6639 out_release_open_seqid:
6640         nfs_release_seqid(data->arg.open_seqid);
6641 out_release_lock_seqid:
6642         nfs_release_seqid(data->arg.lock_seqid);
6643 out_wait:
6644         nfs4_sequence_done(task, &data->res.seq_res);
6645         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6646 }
6647
6648 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6649 {
6650         struct nfs4_lockdata *data = calldata;
6651         struct nfs4_lock_state *lsp = data->lsp;
6652
6653         dprintk("%s: begin!\n", __func__);
6654
6655         if (!nfs4_sequence_done(task, &data->res.seq_res))
6656                 return;
6657
6658         data->rpc_status = task->tk_status;
6659         switch (task->tk_status) {
6660         case 0:
6661                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6662                                 data->timestamp);
6663                 if (data->arg.new_lock && !data->cancelled) {
6664                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6665                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6666                                 goto out_restart;
6667                 }
6668                 if (data->arg.new_lock_owner != 0) {
6669                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6670                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6671                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6672                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6673                         goto out_restart;
6674                 break;
6675         case -NFS4ERR_BAD_STATEID:
6676         case -NFS4ERR_OLD_STATEID:
6677         case -NFS4ERR_STALE_STATEID:
6678         case -NFS4ERR_EXPIRED:
6679                 if (data->arg.new_lock_owner != 0) {
6680                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6681                                                 &lsp->ls_state->open_stateid))
6682                                 goto out_restart;
6683                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6684                                                 &lsp->ls_stateid))
6685                                 goto out_restart;
6686         }
6687 out_done:
6688         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6689         return;
6690 out_restart:
6691         if (!data->cancelled)
6692                 rpc_restart_call_prepare(task);
6693         goto out_done;
6694 }
6695
6696 static void nfs4_lock_release(void *calldata)
6697 {
6698         struct nfs4_lockdata *data = calldata;
6699
6700         dprintk("%s: begin!\n", __func__);
6701         nfs_free_seqid(data->arg.open_seqid);
6702         if (data->cancelled && data->rpc_status == 0) {
6703                 struct rpc_task *task;
6704                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6705                                 data->arg.lock_seqid);
6706                 if (!IS_ERR(task))
6707                         rpc_put_task_async(task);
6708                 dprintk("%s: cancelling lock!\n", __func__);
6709         } else
6710                 nfs_free_seqid(data->arg.lock_seqid);
6711         nfs4_put_lock_state(data->lsp);
6712         put_nfs_open_context(data->ctx);
6713         kfree(data);
6714         dprintk("%s: done!\n", __func__);
6715 }
6716
6717 static const struct rpc_call_ops nfs4_lock_ops = {
6718         .rpc_call_prepare = nfs4_lock_prepare,
6719         .rpc_call_done = nfs4_lock_done,
6720         .rpc_release = nfs4_lock_release,
6721 };
6722
6723 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6724 {
6725         switch (error) {
6726         case -NFS4ERR_ADMIN_REVOKED:
6727         case -NFS4ERR_EXPIRED:
6728         case -NFS4ERR_BAD_STATEID:
6729                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6730                 if (new_lock_owner != 0 ||
6731                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6732                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6733                 break;
6734         case -NFS4ERR_STALE_STATEID:
6735                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6736                 nfs4_schedule_lease_recovery(server->nfs_client);
6737         };
6738 }
6739
6740 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6741 {
6742         struct nfs4_lockdata *data;
6743         struct rpc_task *task;
6744         struct rpc_message msg = {
6745                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6746                 .rpc_cred = state->owner->so_cred,
6747         };
6748         struct rpc_task_setup task_setup_data = {
6749                 .rpc_client = NFS_CLIENT(state->inode),
6750                 .rpc_message = &msg,
6751                 .callback_ops = &nfs4_lock_ops,
6752                 .workqueue = nfsiod_workqueue,
6753                 .flags = RPC_TASK_ASYNC,
6754         };
6755         int ret;
6756
6757         dprintk("%s: begin!\n", __func__);
6758         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6759                         fl->fl_u.nfs4_fl.owner,
6760                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6761         if (data == NULL)
6762                 return -ENOMEM;
6763         if (IS_SETLKW(cmd))
6764                 data->arg.block = 1;
6765         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
6766                                 recovery_type > NFS_LOCK_NEW);
6767         msg.rpc_argp = &data->arg;
6768         msg.rpc_resp = &data->res;
6769         task_setup_data.callback_data = data;
6770         if (recovery_type > NFS_LOCK_NEW) {
6771                 if (recovery_type == NFS_LOCK_RECLAIM)
6772                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6773         } else
6774                 data->arg.new_lock = 1;
6775         task = rpc_run_task(&task_setup_data);
6776         if (IS_ERR(task))
6777                 return PTR_ERR(task);
6778         ret = rpc_wait_for_completion_task(task);
6779         if (ret == 0) {
6780                 ret = data->rpc_status;
6781                 if (ret)
6782                         nfs4_handle_setlk_error(data->server, data->lsp,
6783                                         data->arg.new_lock_owner, ret);
6784         } else
6785                 data->cancelled = true;
6786         rpc_put_task(task);
6787         dprintk("%s: done, ret = %d!\n", __func__, ret);
6788         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6789         return ret;
6790 }
6791
6792 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6793 {
6794         struct nfs_server *server = NFS_SERVER(state->inode);
6795         struct nfs4_exception exception = {
6796                 .inode = state->inode,
6797         };
6798         int err;
6799
6800         do {
6801                 /* Cache the lock if possible... */
6802                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6803                         return 0;
6804                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6805                 if (err != -NFS4ERR_DELAY)
6806                         break;
6807                 nfs4_handle_exception(server, err, &exception);
6808         } while (exception.retry);
6809         return err;
6810 }
6811
6812 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6813 {
6814         struct nfs_server *server = NFS_SERVER(state->inode);
6815         struct nfs4_exception exception = {
6816                 .inode = state->inode,
6817         };
6818         int err;
6819
6820         err = nfs4_set_lock_state(state, request);
6821         if (err != 0)
6822                 return err;
6823         if (!recover_lost_locks) {
6824                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6825                 return 0;
6826         }
6827         do {
6828                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6829                         return 0;
6830                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6831                 switch (err) {
6832                 default:
6833                         goto out;
6834                 case -NFS4ERR_GRACE:
6835                 case -NFS4ERR_DELAY:
6836                         nfs4_handle_exception(server, err, &exception);
6837                         err = 0;
6838                 }
6839         } while (exception.retry);
6840 out:
6841         return err;
6842 }
6843
6844 #if defined(CONFIG_NFS_V4_1)
6845 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6846 {
6847         struct nfs4_lock_state *lsp;
6848         int status;
6849
6850         status = nfs4_set_lock_state(state, request);
6851         if (status != 0)
6852                 return status;
6853         lsp = request->fl_u.nfs4_fl.owner;
6854         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6855             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6856                 return 0;
6857         return nfs4_lock_expired(state, request);
6858 }
6859 #endif
6860
6861 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6862 {
6863         struct nfs_inode *nfsi = NFS_I(state->inode);
6864         struct nfs4_state_owner *sp = state->owner;
6865         unsigned char fl_flags = request->fl_flags;
6866         int status;
6867
6868         request->fl_flags |= FL_ACCESS;
6869         status = locks_lock_inode_wait(state->inode, request);
6870         if (status < 0)
6871                 goto out;
6872         mutex_lock(&sp->so_delegreturn_mutex);
6873         down_read(&nfsi->rwsem);
6874         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6875                 /* Yes: cache locks! */
6876                 /* ...but avoid races with delegation recall... */
6877                 request->fl_flags = fl_flags & ~FL_SLEEP;
6878                 status = locks_lock_inode_wait(state->inode, request);
6879                 up_read(&nfsi->rwsem);
6880                 mutex_unlock(&sp->so_delegreturn_mutex);
6881                 goto out;
6882         }
6883         up_read(&nfsi->rwsem);
6884         mutex_unlock(&sp->so_delegreturn_mutex);
6885         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6886 out:
6887         request->fl_flags = fl_flags;
6888         return status;
6889 }
6890
6891 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6892 {
6893         struct nfs4_exception exception = {
6894                 .state = state,
6895                 .inode = state->inode,
6896                 .interruptible = true,
6897         };
6898         int err;
6899
6900         do {
6901                 err = _nfs4_proc_setlk(state, cmd, request);
6902                 if (err == -NFS4ERR_DENIED)
6903                         err = -EAGAIN;
6904                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6905                                 err, &exception);
6906         } while (exception.retry);
6907         return err;
6908 }
6909
6910 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6911 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6912
6913 static int
6914 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6915                         struct file_lock *request)
6916 {
6917         int             status = -ERESTARTSYS;
6918         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6919
6920         while(!signalled()) {
6921                 status = nfs4_proc_setlk(state, cmd, request);
6922                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6923                         break;
6924                 freezable_schedule_timeout_interruptible(timeout);
6925                 timeout *= 2;
6926                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6927                 status = -ERESTARTSYS;
6928         }
6929         return status;
6930 }
6931
6932 #ifdef CONFIG_NFS_V4_1
6933 struct nfs4_lock_waiter {
6934         struct task_struct      *task;
6935         struct inode            *inode;
6936         struct nfs_lowner       *owner;
6937 };
6938
6939 static int
6940 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
6941 {
6942         int ret;
6943         struct nfs4_lock_waiter *waiter = wait->private;
6944
6945         /* NULL key means to wake up everyone */
6946         if (key) {
6947                 struct cb_notify_lock_args      *cbnl = key;
6948                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
6949                                                 *wowner = waiter->owner;
6950
6951                 /* Only wake if the callback was for the same owner. */
6952                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
6953                         return 0;
6954
6955                 /* Make sure it's for the right inode */
6956                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6957                         return 0;
6958         }
6959
6960         /* override "private" so we can use default_wake_function */
6961         wait->private = waiter->task;
6962         ret = woken_wake_function(wait, mode, flags, key);
6963         if (ret)
6964                 list_del_init(&wait->entry);
6965         wait->private = waiter;
6966         return ret;
6967 }
6968
6969 static int
6970 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6971 {
6972         int status = -ERESTARTSYS;
6973         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6974         struct nfs_server *server = NFS_SERVER(state->inode);
6975         struct nfs_client *clp = server->nfs_client;
6976         wait_queue_head_t *q = &clp->cl_lock_waitq;
6977         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6978                                     .id = lsp->ls_seqid.owner_id,
6979                                     .s_dev = server->s_dev };
6980         struct nfs4_lock_waiter waiter = { .task  = current,
6981                                            .inode = state->inode,
6982                                            .owner = &owner};
6983         wait_queue_entry_t wait;
6984
6985         /* Don't bother with waitqueue if we don't expect a callback */
6986         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6987                 return nfs4_retry_setlk_simple(state, cmd, request);
6988
6989         init_wait(&wait);
6990         wait.private = &waiter;
6991         wait.func = nfs4_wake_lock_waiter;
6992
6993         while(!signalled()) {
6994                 add_wait_queue(q, &wait);
6995                 status = nfs4_proc_setlk(state, cmd, request);
6996                 if ((status != -EAGAIN) || IS_SETLK(cmd)) {
6997                         finish_wait(q, &wait);
6998                         break;
6999                 }
7000
7001                 status = -ERESTARTSYS;
7002                 freezer_do_not_count();
7003                 wait_woken(&wait, TASK_INTERRUPTIBLE, NFS4_LOCK_MAXTIMEOUT);
7004                 freezer_count();
7005                 finish_wait(q, &wait);
7006         }
7007
7008         return status;
7009 }
7010 #else /* !CONFIG_NFS_V4_1 */
7011 static inline int
7012 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7013 {
7014         return nfs4_retry_setlk_simple(state, cmd, request);
7015 }
7016 #endif
7017
7018 static int
7019 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7020 {
7021         struct nfs_open_context *ctx;
7022         struct nfs4_state *state;
7023         int status;
7024
7025         /* verify open state */
7026         ctx = nfs_file_open_context(filp);
7027         state = ctx->state;
7028
7029         if (IS_GETLK(cmd)) {
7030                 if (state != NULL)
7031                         return nfs4_proc_getlk(state, F_GETLK, request);
7032                 return 0;
7033         }
7034
7035         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7036                 return -EINVAL;
7037
7038         if (request->fl_type == F_UNLCK) {
7039                 if (state != NULL)
7040                         return nfs4_proc_unlck(state, cmd, request);
7041                 return 0;
7042         }
7043
7044         if (state == NULL)
7045                 return -ENOLCK;
7046
7047         if ((request->fl_flags & FL_POSIX) &&
7048             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7049                 return -ENOLCK;
7050
7051         /*
7052          * Don't rely on the VFS having checked the file open mode,
7053          * since it won't do this for flock() locks.
7054          */
7055         switch (request->fl_type) {
7056         case F_RDLCK:
7057                 if (!(filp->f_mode & FMODE_READ))
7058                         return -EBADF;
7059                 break;
7060         case F_WRLCK:
7061                 if (!(filp->f_mode & FMODE_WRITE))
7062                         return -EBADF;
7063         }
7064
7065         status = nfs4_set_lock_state(state, request);
7066         if (status != 0)
7067                 return status;
7068
7069         return nfs4_retry_setlk(state, cmd, request);
7070 }
7071
7072 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7073 {
7074         struct nfs_server *server = NFS_SERVER(state->inode);
7075         int err;
7076
7077         err = nfs4_set_lock_state(state, fl);
7078         if (err != 0)
7079                 return err;
7080         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7081         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7082 }
7083
7084 struct nfs_release_lockowner_data {
7085         struct nfs4_lock_state *lsp;
7086         struct nfs_server *server;
7087         struct nfs_release_lockowner_args args;
7088         struct nfs_release_lockowner_res res;
7089         unsigned long timestamp;
7090 };
7091
7092 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7093 {
7094         struct nfs_release_lockowner_data *data = calldata;
7095         struct nfs_server *server = data->server;
7096         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7097                            &data->res.seq_res, task);
7098         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7099         data->timestamp = jiffies;
7100 }
7101
7102 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7103 {
7104         struct nfs_release_lockowner_data *data = calldata;
7105         struct nfs_server *server = data->server;
7106
7107         nfs40_sequence_done(task, &data->res.seq_res);
7108
7109         switch (task->tk_status) {
7110         case 0:
7111                 renew_lease(server, data->timestamp);
7112                 break;
7113         case -NFS4ERR_STALE_CLIENTID:
7114         case -NFS4ERR_EXPIRED:
7115                 nfs4_schedule_lease_recovery(server->nfs_client);
7116                 break;
7117         case -NFS4ERR_LEASE_MOVED:
7118         case -NFS4ERR_DELAY:
7119                 if (nfs4_async_handle_error(task, server,
7120                                             NULL, NULL) == -EAGAIN)
7121                         rpc_restart_call_prepare(task);
7122         }
7123 }
7124
7125 static void nfs4_release_lockowner_release(void *calldata)
7126 {
7127         struct nfs_release_lockowner_data *data = calldata;
7128         nfs4_free_lock_state(data->server, data->lsp);
7129         kfree(calldata);
7130 }
7131
7132 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7133         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7134         .rpc_call_done = nfs4_release_lockowner_done,
7135         .rpc_release = nfs4_release_lockowner_release,
7136 };
7137
7138 static void
7139 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7140 {
7141         struct nfs_release_lockowner_data *data;
7142         struct rpc_message msg = {
7143                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7144         };
7145
7146         if (server->nfs_client->cl_mvops->minor_version != 0)
7147                 return;
7148
7149         data = kmalloc(sizeof(*data), GFP_NOFS);
7150         if (!data)
7151                 return;
7152         data->lsp = lsp;
7153         data->server = server;
7154         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7155         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7156         data->args.lock_owner.s_dev = server->s_dev;
7157
7158         msg.rpc_argp = &data->args;
7159         msg.rpc_resp = &data->res;
7160         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7161         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7162 }
7163
7164 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7165
7166 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7167                                    struct dentry *unused, struct inode *inode,
7168                                    const char *key, const void *buf,
7169                                    size_t buflen, int flags)
7170 {
7171         return nfs4_proc_set_acl(inode, buf, buflen);
7172 }
7173
7174 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7175                                    struct dentry *unused, struct inode *inode,
7176                                    const char *key, void *buf, size_t buflen)
7177 {
7178         return nfs4_proc_get_acl(inode, buf, buflen);
7179 }
7180
7181 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7182 {
7183         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7184 }
7185
7186 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7187
7188 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7189                                      struct dentry *unused, struct inode *inode,
7190                                      const char *key, const void *buf,
7191                                      size_t buflen, int flags)
7192 {
7193         if (security_ismaclabel(key))
7194                 return nfs4_set_security_label(inode, buf, buflen);
7195
7196         return -EOPNOTSUPP;
7197 }
7198
7199 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7200                                      struct dentry *unused, struct inode *inode,
7201                                      const char *key, void *buf, size_t buflen)
7202 {
7203         if (security_ismaclabel(key))
7204                 return nfs4_get_security_label(inode, buf, buflen);
7205         return -EOPNOTSUPP;
7206 }
7207
7208 static ssize_t
7209 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7210 {
7211         int len = 0;
7212
7213         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7214                 len = security_inode_listsecurity(inode, list, list_len);
7215                 if (list_len && len > list_len)
7216                         return -ERANGE;
7217         }
7218         return len;
7219 }
7220
7221 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7222         .prefix = XATTR_SECURITY_PREFIX,
7223         .get    = nfs4_xattr_get_nfs4_label,
7224         .set    = nfs4_xattr_set_nfs4_label,
7225 };
7226
7227 #else
7228
7229 static ssize_t
7230 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7231 {
7232         return 0;
7233 }
7234
7235 #endif
7236
7237 /*
7238  * nfs_fhget will use either the mounted_on_fileid or the fileid
7239  */
7240 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7241 {
7242         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7243                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7244               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7245               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7246                 return;
7247
7248         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7249                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7250         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7251         fattr->nlink = 2;
7252 }
7253
7254 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7255                                    const struct qstr *name,
7256                                    struct nfs4_fs_locations *fs_locations,
7257                                    struct page *page)
7258 {
7259         struct nfs_server *server = NFS_SERVER(dir);
7260         u32 bitmask[3];
7261         struct nfs4_fs_locations_arg args = {
7262                 .dir_fh = NFS_FH(dir),
7263                 .name = name,
7264                 .page = page,
7265                 .bitmask = bitmask,
7266         };
7267         struct nfs4_fs_locations_res res = {
7268                 .fs_locations = fs_locations,
7269         };
7270         struct rpc_message msg = {
7271                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7272                 .rpc_argp = &args,
7273                 .rpc_resp = &res,
7274         };
7275         int status;
7276
7277         dprintk("%s: start\n", __func__);
7278
7279         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7280         bitmask[1] = nfs4_fattr_bitmap[1];
7281
7282         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7283          * is not supported */
7284         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7285                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7286         else
7287                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7288
7289         nfs_fattr_init(&fs_locations->fattr);
7290         fs_locations->server = server;
7291         fs_locations->nlocations = 0;
7292         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7293         dprintk("%s: returned status = %d\n", __func__, status);
7294         return status;
7295 }
7296
7297 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7298                            const struct qstr *name,
7299                            struct nfs4_fs_locations *fs_locations,
7300                            struct page *page)
7301 {
7302         struct nfs4_exception exception = {
7303                 .interruptible = true,
7304         };
7305         int err;
7306         do {
7307                 err = _nfs4_proc_fs_locations(client, dir, name,
7308                                 fs_locations, page);
7309                 trace_nfs4_get_fs_locations(dir, name, err);
7310                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7311                                 &exception);
7312         } while (exception.retry);
7313         return err;
7314 }
7315
7316 /*
7317  * This operation also signals the server that this client is
7318  * performing migration recovery.  The server can stop returning
7319  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7320  * appended to this compound to identify the client ID which is
7321  * performing recovery.
7322  */
7323 static int _nfs40_proc_get_locations(struct inode *inode,
7324                                      struct nfs4_fs_locations *locations,
7325                                      struct page *page, const struct cred *cred)
7326 {
7327         struct nfs_server *server = NFS_SERVER(inode);
7328         struct rpc_clnt *clnt = server->client;
7329         u32 bitmask[2] = {
7330                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7331         };
7332         struct nfs4_fs_locations_arg args = {
7333                 .clientid       = server->nfs_client->cl_clientid,
7334                 .fh             = NFS_FH(inode),
7335                 .page           = page,
7336                 .bitmask        = bitmask,
7337                 .migration      = 1,            /* skip LOOKUP */
7338                 .renew          = 1,            /* append RENEW */
7339         };
7340         struct nfs4_fs_locations_res res = {
7341                 .fs_locations   = locations,
7342                 .migration      = 1,
7343                 .renew          = 1,
7344         };
7345         struct rpc_message msg = {
7346                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7347                 .rpc_argp       = &args,
7348                 .rpc_resp       = &res,
7349                 .rpc_cred       = cred,
7350         };
7351         unsigned long now = jiffies;
7352         int status;
7353
7354         nfs_fattr_init(&locations->fattr);
7355         locations->server = server;
7356         locations->nlocations = 0;
7357
7358         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7359         status = nfs4_call_sync_sequence(clnt, server, &msg,
7360                                         &args.seq_args, &res.seq_res);
7361         if (status)
7362                 return status;
7363
7364         renew_lease(server, now);
7365         return 0;
7366 }
7367
7368 #ifdef CONFIG_NFS_V4_1
7369
7370 /*
7371  * This operation also signals the server that this client is
7372  * performing migration recovery.  The server can stop asserting
7373  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7374  * performing this operation is identified in the SEQUENCE
7375  * operation in this compound.
7376  *
7377  * When the client supports GETATTR(fs_locations_info), it can
7378  * be plumbed in here.
7379  */
7380 static int _nfs41_proc_get_locations(struct inode *inode,
7381                                      struct nfs4_fs_locations *locations,
7382                                      struct page *page, const struct cred *cred)
7383 {
7384         struct nfs_server *server = NFS_SERVER(inode);
7385         struct rpc_clnt *clnt = server->client;
7386         u32 bitmask[2] = {
7387                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7388         };
7389         struct nfs4_fs_locations_arg args = {
7390                 .fh             = NFS_FH(inode),
7391                 .page           = page,
7392                 .bitmask        = bitmask,
7393                 .migration      = 1,            /* skip LOOKUP */
7394         };
7395         struct nfs4_fs_locations_res res = {
7396                 .fs_locations   = locations,
7397                 .migration      = 1,
7398         };
7399         struct rpc_message msg = {
7400                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7401                 .rpc_argp       = &args,
7402                 .rpc_resp       = &res,
7403                 .rpc_cred       = cred,
7404         };
7405         int status;
7406
7407         nfs_fattr_init(&locations->fattr);
7408         locations->server = server;
7409         locations->nlocations = 0;
7410
7411         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7412         status = nfs4_call_sync_sequence(clnt, server, &msg,
7413                                         &args.seq_args, &res.seq_res);
7414         if (status == NFS4_OK &&
7415             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7416                 status = -NFS4ERR_LEASE_MOVED;
7417         return status;
7418 }
7419
7420 #endif  /* CONFIG_NFS_V4_1 */
7421
7422 /**
7423  * nfs4_proc_get_locations - discover locations for a migrated FSID
7424  * @inode: inode on FSID that is migrating
7425  * @locations: result of query
7426  * @page: buffer
7427  * @cred: credential to use for this operation
7428  *
7429  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7430  * operation failed, or a negative errno if a local error occurred.
7431  *
7432  * On success, "locations" is filled in, but if the server has
7433  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7434  * asserted.
7435  *
7436  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7437  * from this client that require migration recovery.
7438  */
7439 int nfs4_proc_get_locations(struct inode *inode,
7440                             struct nfs4_fs_locations *locations,
7441                             struct page *page, const struct cred *cred)
7442 {
7443         struct nfs_server *server = NFS_SERVER(inode);
7444         struct nfs_client *clp = server->nfs_client;
7445         const struct nfs4_mig_recovery_ops *ops =
7446                                         clp->cl_mvops->mig_recovery_ops;
7447         struct nfs4_exception exception = {
7448                 .interruptible = true,
7449         };
7450         int status;
7451
7452         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7453                 (unsigned long long)server->fsid.major,
7454                 (unsigned long long)server->fsid.minor,
7455                 clp->cl_hostname);
7456         nfs_display_fhandle(NFS_FH(inode), __func__);
7457
7458         do {
7459                 status = ops->get_locations(inode, locations, page, cred);
7460                 if (status != -NFS4ERR_DELAY)
7461                         break;
7462                 nfs4_handle_exception(server, status, &exception);
7463         } while (exception.retry);
7464         return status;
7465 }
7466
7467 /*
7468  * This operation also signals the server that this client is
7469  * performing "lease moved" recovery.  The server can stop
7470  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7471  * is appended to this compound to identify the client ID which is
7472  * performing recovery.
7473  */
7474 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7475 {
7476         struct nfs_server *server = NFS_SERVER(inode);
7477         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7478         struct rpc_clnt *clnt = server->client;
7479         struct nfs4_fsid_present_arg args = {
7480                 .fh             = NFS_FH(inode),
7481                 .clientid       = clp->cl_clientid,
7482                 .renew          = 1,            /* append RENEW */
7483         };
7484         struct nfs4_fsid_present_res res = {
7485                 .renew          = 1,
7486         };
7487         struct rpc_message msg = {
7488                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7489                 .rpc_argp       = &args,
7490                 .rpc_resp       = &res,
7491                 .rpc_cred       = cred,
7492         };
7493         unsigned long now = jiffies;
7494         int status;
7495
7496         res.fh = nfs_alloc_fhandle();
7497         if (res.fh == NULL)
7498                 return -ENOMEM;
7499
7500         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7501         status = nfs4_call_sync_sequence(clnt, server, &msg,
7502                                                 &args.seq_args, &res.seq_res);
7503         nfs_free_fhandle(res.fh);
7504         if (status)
7505                 return status;
7506
7507         do_renew_lease(clp, now);
7508         return 0;
7509 }
7510
7511 #ifdef CONFIG_NFS_V4_1
7512
7513 /*
7514  * This operation also signals the server that this client is
7515  * performing "lease moved" recovery.  The server can stop asserting
7516  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7517  * this operation is identified in the SEQUENCE operation in this
7518  * compound.
7519  */
7520 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
7521 {
7522         struct nfs_server *server = NFS_SERVER(inode);
7523         struct rpc_clnt *clnt = server->client;
7524         struct nfs4_fsid_present_arg args = {
7525                 .fh             = NFS_FH(inode),
7526         };
7527         struct nfs4_fsid_present_res res = {
7528         };
7529         struct rpc_message msg = {
7530                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7531                 .rpc_argp       = &args,
7532                 .rpc_resp       = &res,
7533                 .rpc_cred       = cred,
7534         };
7535         int status;
7536
7537         res.fh = nfs_alloc_fhandle();
7538         if (res.fh == NULL)
7539                 return -ENOMEM;
7540
7541         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7542         status = nfs4_call_sync_sequence(clnt, server, &msg,
7543                                                 &args.seq_args, &res.seq_res);
7544         nfs_free_fhandle(res.fh);
7545         if (status == NFS4_OK &&
7546             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7547                 status = -NFS4ERR_LEASE_MOVED;
7548         return status;
7549 }
7550
7551 #endif  /* CONFIG_NFS_V4_1 */
7552
7553 /**
7554  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7555  * @inode: inode on FSID to check
7556  * @cred: credential to use for this operation
7557  *
7558  * Server indicates whether the FSID is present, moved, or not
7559  * recognized.  This operation is necessary to clear a LEASE_MOVED
7560  * condition for this client ID.
7561  *
7562  * Returns NFS4_OK if the FSID is present on this server,
7563  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7564  *  NFS4ERR code if some error occurred on the server, or a
7565  *  negative errno if a local failure occurred.
7566  */
7567 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
7568 {
7569         struct nfs_server *server = NFS_SERVER(inode);
7570         struct nfs_client *clp = server->nfs_client;
7571         const struct nfs4_mig_recovery_ops *ops =
7572                                         clp->cl_mvops->mig_recovery_ops;
7573         struct nfs4_exception exception = {
7574                 .interruptible = true,
7575         };
7576         int status;
7577
7578         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7579                 (unsigned long long)server->fsid.major,
7580                 (unsigned long long)server->fsid.minor,
7581                 clp->cl_hostname);
7582         nfs_display_fhandle(NFS_FH(inode), __func__);
7583
7584         do {
7585                 status = ops->fsid_present(inode, cred);
7586                 if (status != -NFS4ERR_DELAY)
7587                         break;
7588                 nfs4_handle_exception(server, status, &exception);
7589         } while (exception.retry);
7590         return status;
7591 }
7592
7593 /*
7594  * If 'use_integrity' is true and the state managment nfs_client
7595  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7596  * and the machine credential as per RFC3530bis and RFC5661 Security
7597  * Considerations sections. Otherwise, just use the user cred with the
7598  * filesystem's rpc_client.
7599  */
7600 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7601 {
7602         int status;
7603         struct nfs4_secinfo_arg args = {
7604                 .dir_fh = NFS_FH(dir),
7605                 .name   = name,
7606         };
7607         struct nfs4_secinfo_res res = {
7608                 .flavors     = flavors,
7609         };
7610         struct rpc_message msg = {
7611                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7612                 .rpc_argp = &args,
7613                 .rpc_resp = &res,
7614         };
7615         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7616         const struct cred *cred = NULL;
7617
7618         if (use_integrity) {
7619                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7620                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7621                 msg.rpc_cred = cred;
7622         }
7623
7624         dprintk("NFS call  secinfo %s\n", name->name);
7625
7626         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7627                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7628
7629         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7630                                 &res.seq_res, 0);
7631         dprintk("NFS reply  secinfo: %d\n", status);
7632
7633         put_cred(cred);
7634
7635         return status;
7636 }
7637
7638 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7639                       struct nfs4_secinfo_flavors *flavors)
7640 {
7641         struct nfs4_exception exception = {
7642                 .interruptible = true,
7643         };
7644         int err;
7645         do {
7646                 err = -NFS4ERR_WRONGSEC;
7647
7648                 /* try to use integrity protection with machine cred */
7649                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7650                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7651
7652                 /*
7653                  * if unable to use integrity protection, or SECINFO with
7654                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7655                  * disallowed by spec, but exists in deployed servers) use
7656                  * the current filesystem's rpc_client and the user cred.
7657                  */
7658                 if (err == -NFS4ERR_WRONGSEC)
7659                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7660
7661                 trace_nfs4_secinfo(dir, name, err);
7662                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7663                                 &exception);
7664         } while (exception.retry);
7665         return err;
7666 }
7667
7668 #ifdef CONFIG_NFS_V4_1
7669 /*
7670  * Check the exchange flags returned by the server for invalid flags, having
7671  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7672  * DS flags set.
7673  */
7674 static int nfs4_check_cl_exchange_flags(u32 flags)
7675 {
7676         if (flags & ~EXCHGID4_FLAG_MASK_R)
7677                 goto out_inval;
7678         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7679             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7680                 goto out_inval;
7681         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7682                 goto out_inval;
7683         return NFS_OK;
7684 out_inval:
7685         return -NFS4ERR_INVAL;
7686 }
7687
7688 static bool
7689 nfs41_same_server_scope(struct nfs41_server_scope *a,
7690                         struct nfs41_server_scope *b)
7691 {
7692         if (a->server_scope_sz != b->server_scope_sz)
7693                 return false;
7694         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7695 }
7696
7697 static void
7698 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7699 {
7700 }
7701
7702 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7703         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7704 };
7705
7706 /*
7707  * nfs4_proc_bind_one_conn_to_session()
7708  *
7709  * The 4.1 client currently uses the same TCP connection for the
7710  * fore and backchannel.
7711  */
7712 static
7713 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7714                 struct rpc_xprt *xprt,
7715                 struct nfs_client *clp,
7716                 const struct cred *cred)
7717 {
7718         int status;
7719         struct nfs41_bind_conn_to_session_args args = {
7720                 .client = clp,
7721                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7722         };
7723         struct nfs41_bind_conn_to_session_res res;
7724         struct rpc_message msg = {
7725                 .rpc_proc =
7726                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7727                 .rpc_argp = &args,
7728                 .rpc_resp = &res,
7729                 .rpc_cred = cred,
7730         };
7731         struct rpc_task_setup task_setup_data = {
7732                 .rpc_client = clnt,
7733                 .rpc_xprt = xprt,
7734                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7735                 .rpc_message = &msg,
7736                 .flags = RPC_TASK_TIMEOUT,
7737         };
7738         struct rpc_task *task;
7739
7740         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7741         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7742                 args.dir = NFS4_CDFC4_FORE;
7743
7744         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7745         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7746                 args.dir = NFS4_CDFC4_FORE;
7747
7748         task = rpc_run_task(&task_setup_data);
7749         if (!IS_ERR(task)) {
7750                 status = task->tk_status;
7751                 rpc_put_task(task);
7752         } else
7753                 status = PTR_ERR(task);
7754         trace_nfs4_bind_conn_to_session(clp, status);
7755         if (status == 0) {
7756                 if (memcmp(res.sessionid.data,
7757                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7758                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7759                         return -EIO;
7760                 }
7761                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7762                         dprintk("NFS: %s: Unexpected direction from server\n",
7763                                 __func__);
7764                         return -EIO;
7765                 }
7766                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7767                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7768                                 __func__);
7769                         return -EIO;
7770                 }
7771         }
7772
7773         return status;
7774 }
7775
7776 struct rpc_bind_conn_calldata {
7777         struct nfs_client *clp;
7778         const struct cred *cred;
7779 };
7780
7781 static int
7782 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7783                 struct rpc_xprt *xprt,
7784                 void *calldata)
7785 {
7786         struct rpc_bind_conn_calldata *p = calldata;
7787
7788         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7789 }
7790
7791 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
7792 {
7793         struct rpc_bind_conn_calldata data = {
7794                 .clp = clp,
7795                 .cred = cred,
7796         };
7797         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7798                         nfs4_proc_bind_conn_to_session_callback, &data);
7799 }
7800
7801 /*
7802  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7803  * and operations we'd like to see to enable certain features in the allow map
7804  */
7805 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7806         .how = SP4_MACH_CRED,
7807         .enforce.u.words = {
7808                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7809                       1 << (OP_EXCHANGE_ID - 32) |
7810                       1 << (OP_CREATE_SESSION - 32) |
7811                       1 << (OP_DESTROY_SESSION - 32) |
7812                       1 << (OP_DESTROY_CLIENTID - 32)
7813         },
7814         .allow.u.words = {
7815                 [0] = 1 << (OP_CLOSE) |
7816                       1 << (OP_OPEN_DOWNGRADE) |
7817                       1 << (OP_LOCKU) |
7818                       1 << (OP_DELEGRETURN) |
7819                       1 << (OP_COMMIT),
7820                 [1] = 1 << (OP_SECINFO - 32) |
7821                       1 << (OP_SECINFO_NO_NAME - 32) |
7822                       1 << (OP_LAYOUTRETURN - 32) |
7823                       1 << (OP_TEST_STATEID - 32) |
7824                       1 << (OP_FREE_STATEID - 32) |
7825                       1 << (OP_WRITE - 32)
7826         }
7827 };
7828
7829 /*
7830  * Select the state protection mode for client `clp' given the server results
7831  * from exchange_id in `sp'.
7832  *
7833  * Returns 0 on success, negative errno otherwise.
7834  */
7835 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7836                                  struct nfs41_state_protection *sp)
7837 {
7838         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7839                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7840                       1 << (OP_EXCHANGE_ID - 32) |
7841                       1 << (OP_CREATE_SESSION - 32) |
7842                       1 << (OP_DESTROY_SESSION - 32) |
7843                       1 << (OP_DESTROY_CLIENTID - 32)
7844         };
7845         unsigned long flags = 0;
7846         unsigned int i;
7847         int ret = 0;
7848
7849         if (sp->how == SP4_MACH_CRED) {
7850                 /* Print state protect result */
7851                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7852                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7853                         if (test_bit(i, sp->enforce.u.longs))
7854                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7855                         if (test_bit(i, sp->allow.u.longs))
7856                                 dfprintk(MOUNT, "  allow op %d\n", i);
7857                 }
7858
7859                 /* make sure nothing is on enforce list that isn't supported */
7860                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7861                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7862                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7863                                 ret = -EINVAL;
7864                                 goto out;
7865                         }
7866                 }
7867
7868                 /*
7869                  * Minimal mode - state operations are allowed to use machine
7870                  * credential.  Note this already happens by default, so the
7871                  * client doesn't have to do anything more than the negotiation.
7872                  *
7873                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7874                  *       we're already using the machine cred for exchange_id
7875                  *       and will never use a different cred.
7876                  */
7877                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7878                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7879                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7880                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7881                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7882                         dfprintk(MOUNT, "  minimal mode enabled\n");
7883                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
7884                 } else {
7885                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7886                         ret = -EINVAL;
7887                         goto out;
7888                 }
7889
7890                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7891                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7892                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7893                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7894                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7895                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
7896                 }
7897
7898                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7899                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7900                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
7901                 }
7902
7903                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7904                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7905                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7906                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
7907                 }
7908
7909                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7910                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7911                         dfprintk(MOUNT, "  stateid mode enabled\n");
7912                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
7913                 }
7914
7915                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7916                         dfprintk(MOUNT, "  write mode enabled\n");
7917                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
7918                 }
7919
7920                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7921                         dfprintk(MOUNT, "  commit mode enabled\n");
7922                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
7923                 }
7924         }
7925 out:
7926         clp->cl_sp4_flags = flags;
7927         return ret;
7928 }
7929
7930 struct nfs41_exchange_id_data {
7931         struct nfs41_exchange_id_res res;
7932         struct nfs41_exchange_id_args args;
7933 };
7934
7935 static void nfs4_exchange_id_release(void *data)
7936 {
7937         struct nfs41_exchange_id_data *cdata =
7938                                         (struct nfs41_exchange_id_data *)data;
7939
7940         nfs_put_client(cdata->args.client);
7941         kfree(cdata->res.impl_id);
7942         kfree(cdata->res.server_scope);
7943         kfree(cdata->res.server_owner);
7944         kfree(cdata);
7945 }
7946
7947 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7948         .rpc_release = nfs4_exchange_id_release,
7949 };
7950
7951 /*
7952  * _nfs4_proc_exchange_id()
7953  *
7954  * Wrapper for EXCHANGE_ID operation.
7955  */
7956 static struct rpc_task *
7957 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
7958                         u32 sp4_how, struct rpc_xprt *xprt)
7959 {
7960         struct rpc_message msg = {
7961                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7962                 .rpc_cred = cred,
7963         };
7964         struct rpc_task_setup task_setup_data = {
7965                 .rpc_client = clp->cl_rpcclient,
7966                 .callback_ops = &nfs4_exchange_id_call_ops,
7967                 .rpc_message = &msg,
7968                 .flags = RPC_TASK_TIMEOUT,
7969         };
7970         struct nfs41_exchange_id_data *calldata;
7971         int status;
7972
7973         if (!refcount_inc_not_zero(&clp->cl_count))
7974                 return ERR_PTR(-EIO);
7975
7976         status = -ENOMEM;
7977         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7978         if (!calldata)
7979                 goto out;
7980
7981         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
7982
7983         status = nfs4_init_uniform_client_string(clp);
7984         if (status)
7985                 goto out_calldata;
7986
7987         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7988                                                 GFP_NOFS);
7989         status = -ENOMEM;
7990         if (unlikely(calldata->res.server_owner == NULL))
7991                 goto out_calldata;
7992
7993         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7994                                         GFP_NOFS);
7995         if (unlikely(calldata->res.server_scope == NULL))
7996                 goto out_server_owner;
7997
7998         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7999         if (unlikely(calldata->res.impl_id == NULL))
8000                 goto out_server_scope;
8001
8002         switch (sp4_how) {
8003         case SP4_NONE:
8004                 calldata->args.state_protect.how = SP4_NONE;
8005                 break;
8006
8007         case SP4_MACH_CRED:
8008                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8009                 break;
8010
8011         default:
8012                 /* unsupported! */
8013                 WARN_ON_ONCE(1);
8014                 status = -EINVAL;
8015                 goto out_impl_id;
8016         }
8017         if (xprt) {
8018                 task_setup_data.rpc_xprt = xprt;
8019                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8020                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8021                                 sizeof(calldata->args.verifier.data));
8022         }
8023         calldata->args.client = clp;
8024         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8025         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8026 #ifdef CONFIG_NFS_V4_1_MIGRATION
8027         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8028 #endif
8029         msg.rpc_argp = &calldata->args;
8030         msg.rpc_resp = &calldata->res;
8031         task_setup_data.callback_data = calldata;
8032
8033         return rpc_run_task(&task_setup_data);
8034
8035 out_impl_id:
8036         kfree(calldata->res.impl_id);
8037 out_server_scope:
8038         kfree(calldata->res.server_scope);
8039 out_server_owner:
8040         kfree(calldata->res.server_owner);
8041 out_calldata:
8042         kfree(calldata);
8043 out:
8044         nfs_put_client(clp);
8045         return ERR_PTR(status);
8046 }
8047
8048 /*
8049  * _nfs4_proc_exchange_id()
8050  *
8051  * Wrapper for EXCHANGE_ID operation.
8052  */
8053 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8054                         u32 sp4_how)
8055 {
8056         struct rpc_task *task;
8057         struct nfs41_exchange_id_args *argp;
8058         struct nfs41_exchange_id_res *resp;
8059         int status;
8060
8061         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8062         if (IS_ERR(task))
8063                 return PTR_ERR(task);
8064
8065         argp = task->tk_msg.rpc_argp;
8066         resp = task->tk_msg.rpc_resp;
8067         status = task->tk_status;
8068         if (status  != 0)
8069                 goto out;
8070
8071         status = nfs4_check_cl_exchange_flags(resp->flags);
8072         if (status  != 0)
8073                 goto out;
8074
8075         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8076         if (status != 0)
8077                 goto out;
8078
8079         clp->cl_clientid = resp->clientid;
8080         clp->cl_exchange_flags = resp->flags;
8081         clp->cl_seqid = resp->seqid;
8082         /* Client ID is not confirmed */
8083         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8084                 clear_bit(NFS4_SESSION_ESTABLISHED,
8085                           &clp->cl_session->session_state);
8086
8087         if (clp->cl_serverscope != NULL &&
8088             !nfs41_same_server_scope(clp->cl_serverscope,
8089                                 resp->server_scope)) {
8090                 dprintk("%s: server_scope mismatch detected\n",
8091                         __func__);
8092                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8093         }
8094
8095         swap(clp->cl_serverowner, resp->server_owner);
8096         swap(clp->cl_serverscope, resp->server_scope);
8097         swap(clp->cl_implid, resp->impl_id);
8098
8099         /* Save the EXCHANGE_ID verifier session trunk tests */
8100         memcpy(clp->cl_confirm.data, argp->verifier.data,
8101                sizeof(clp->cl_confirm.data));
8102 out:
8103         trace_nfs4_exchange_id(clp, status);
8104         rpc_put_task(task);
8105         return status;
8106 }
8107
8108 /*
8109  * nfs4_proc_exchange_id()
8110  *
8111  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8112  *
8113  * Since the clientid has expired, all compounds using sessions
8114  * associated with the stale clientid will be returning
8115  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8116  * be in some phase of session reset.
8117  *
8118  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8119  */
8120 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8121 {
8122         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8123         int status;
8124
8125         /* try SP4_MACH_CRED if krb5i/p */
8126         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8127             authflavor == RPC_AUTH_GSS_KRB5P) {
8128                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8129                 if (!status)
8130                         return 0;
8131         }
8132
8133         /* try SP4_NONE */
8134         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8135 }
8136
8137 /**
8138  * nfs4_test_session_trunk
8139  *
8140  * This is an add_xprt_test() test function called from
8141  * rpc_clnt_setup_test_and_add_xprt.
8142  *
8143  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8144  * and is dereferrenced in nfs4_exchange_id_release
8145  *
8146  * Upon success, add the new transport to the rpc_clnt
8147  *
8148  * @clnt: struct rpc_clnt to get new transport
8149  * @xprt: the rpc_xprt to test
8150  * @data: call data for _nfs4_proc_exchange_id.
8151  */
8152 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8153                             void *data)
8154 {
8155         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8156         struct rpc_task *task;
8157         int status;
8158
8159         u32 sp4_how;
8160
8161         dprintk("--> %s try %s\n", __func__,
8162                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8163
8164         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8165
8166         /* Test connection for session trunking. Async exchange_id call */
8167         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8168         if (IS_ERR(task))
8169                 return;
8170
8171         status = task->tk_status;
8172         if (status == 0)
8173                 status = nfs4_detect_session_trunking(adata->clp,
8174                                 task->tk_msg.rpc_resp, xprt);
8175
8176         if (status == 0)
8177                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8178
8179         rpc_put_task(task);
8180 }
8181 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8182
8183 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8184                 const struct cred *cred)
8185 {
8186         struct rpc_message msg = {
8187                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8188                 .rpc_argp = clp,
8189                 .rpc_cred = cred,
8190         };
8191         int status;
8192
8193         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8194         trace_nfs4_destroy_clientid(clp, status);
8195         if (status)
8196                 dprintk("NFS: Got error %d from the server %s on "
8197                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8198         return status;
8199 }
8200
8201 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8202                 const struct cred *cred)
8203 {
8204         unsigned int loop;
8205         int ret;
8206
8207         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8208                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8209                 switch (ret) {
8210                 case -NFS4ERR_DELAY:
8211                 case -NFS4ERR_CLIENTID_BUSY:
8212                         ssleep(1);
8213                         break;
8214                 default:
8215                         return ret;
8216                 }
8217         }
8218         return 0;
8219 }
8220
8221 int nfs4_destroy_clientid(struct nfs_client *clp)
8222 {
8223         const struct cred *cred;
8224         int ret = 0;
8225
8226         if (clp->cl_mvops->minor_version < 1)
8227                 goto out;
8228         if (clp->cl_exchange_flags == 0)
8229                 goto out;
8230         if (clp->cl_preserve_clid)
8231                 goto out;
8232         cred = nfs4_get_clid_cred(clp);
8233         ret = nfs4_proc_destroy_clientid(clp, cred);
8234         put_cred(cred);
8235         switch (ret) {
8236         case 0:
8237         case -NFS4ERR_STALE_CLIENTID:
8238                 clp->cl_exchange_flags = 0;
8239         }
8240 out:
8241         return ret;
8242 }
8243
8244 struct nfs4_get_lease_time_data {
8245         struct nfs4_get_lease_time_args *args;
8246         struct nfs4_get_lease_time_res *res;
8247         struct nfs_client *clp;
8248 };
8249
8250 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8251                                         void *calldata)
8252 {
8253         struct nfs4_get_lease_time_data *data =
8254                         (struct nfs4_get_lease_time_data *)calldata;
8255
8256         dprintk("--> %s\n", __func__);
8257         /* just setup sequence, do not trigger session recovery
8258            since we're invoked within one */
8259         nfs4_setup_sequence(data->clp,
8260                         &data->args->la_seq_args,
8261                         &data->res->lr_seq_res,
8262                         task);
8263         dprintk("<-- %s\n", __func__);
8264 }
8265
8266 /*
8267  * Called from nfs4_state_manager thread for session setup, so don't recover
8268  * from sequence operation or clientid errors.
8269  */
8270 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8271 {
8272         struct nfs4_get_lease_time_data *data =
8273                         (struct nfs4_get_lease_time_data *)calldata;
8274
8275         dprintk("--> %s\n", __func__);
8276         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
8277                 return;
8278         switch (task->tk_status) {
8279         case -NFS4ERR_DELAY:
8280         case -NFS4ERR_GRACE:
8281                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8282                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8283                 task->tk_status = 0;
8284                 /* fall through */
8285         case -NFS4ERR_RETRY_UNCACHED_REP:
8286                 rpc_restart_call_prepare(task);
8287                 return;
8288         }
8289         dprintk("<-- %s\n", __func__);
8290 }
8291
8292 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8293         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8294         .rpc_call_done = nfs4_get_lease_time_done,
8295 };
8296
8297 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8298 {
8299         struct rpc_task *task;
8300         struct nfs4_get_lease_time_args args;
8301         struct nfs4_get_lease_time_res res = {
8302                 .lr_fsinfo = fsinfo,
8303         };
8304         struct nfs4_get_lease_time_data data = {
8305                 .args = &args,
8306                 .res = &res,
8307                 .clp = clp,
8308         };
8309         struct rpc_message msg = {
8310                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8311                 .rpc_argp = &args,
8312                 .rpc_resp = &res,
8313         };
8314         struct rpc_task_setup task_setup = {
8315                 .rpc_client = clp->cl_rpcclient,
8316                 .rpc_message = &msg,
8317                 .callback_ops = &nfs4_get_lease_time_ops,
8318                 .callback_data = &data,
8319                 .flags = RPC_TASK_TIMEOUT,
8320         };
8321         int status;
8322
8323         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8324         task = rpc_run_task(&task_setup);
8325
8326         if (IS_ERR(task))
8327                 return PTR_ERR(task);
8328
8329         status = task->tk_status;
8330         rpc_put_task(task);
8331         return status;
8332 }
8333
8334 /*
8335  * Initialize the values to be used by the client in CREATE_SESSION
8336  * If nfs4_init_session set the fore channel request and response sizes,
8337  * use them.
8338  *
8339  * Set the back channel max_resp_sz_cached to zero to force the client to
8340  * always set csa_cachethis to FALSE because the current implementation
8341  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8342  */
8343 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8344                                     struct rpc_clnt *clnt)
8345 {
8346         unsigned int max_rqst_sz, max_resp_sz;
8347         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8348
8349         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8350         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8351
8352         /* Fore channel attributes */
8353         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8354         args->fc_attrs.max_resp_sz = max_resp_sz;
8355         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8356         args->fc_attrs.max_reqs = max_session_slots;
8357
8358         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8359                 "max_ops=%u max_reqs=%u\n",
8360                 __func__,
8361                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8362                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8363
8364         /* Back channel attributes */
8365         args->bc_attrs.max_rqst_sz = max_bc_payload;
8366         args->bc_attrs.max_resp_sz = max_bc_payload;
8367         args->bc_attrs.max_resp_sz_cached = 0;
8368         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8369         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8370
8371         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8372                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8373                 __func__,
8374                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8375                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8376                 args->bc_attrs.max_reqs);
8377 }
8378
8379 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8380                 struct nfs41_create_session_res *res)
8381 {
8382         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8383         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8384
8385         if (rcvd->max_resp_sz > sent->max_resp_sz)
8386                 return -EINVAL;
8387         /*
8388          * Our requested max_ops is the minimum we need; we're not
8389          * prepared to break up compounds into smaller pieces than that.
8390          * So, no point even trying to continue if the server won't
8391          * cooperate:
8392          */
8393         if (rcvd->max_ops < sent->max_ops)
8394                 return -EINVAL;
8395         if (rcvd->max_reqs == 0)
8396                 return -EINVAL;
8397         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8398                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8399         return 0;
8400 }
8401
8402 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8403                 struct nfs41_create_session_res *res)
8404 {
8405         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8406         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8407
8408         if (!(res->flags & SESSION4_BACK_CHAN))
8409                 goto out;
8410         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8411                 return -EINVAL;
8412         if (rcvd->max_resp_sz < sent->max_resp_sz)
8413                 return -EINVAL;
8414         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8415                 return -EINVAL;
8416         if (rcvd->max_ops > sent->max_ops)
8417                 return -EINVAL;
8418         if (rcvd->max_reqs > sent->max_reqs)
8419                 return -EINVAL;
8420 out:
8421         return 0;
8422 }
8423
8424 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8425                                      struct nfs41_create_session_res *res)
8426 {
8427         int ret;
8428
8429         ret = nfs4_verify_fore_channel_attrs(args, res);
8430         if (ret)
8431                 return ret;
8432         return nfs4_verify_back_channel_attrs(args, res);
8433 }
8434
8435 static void nfs4_update_session(struct nfs4_session *session,
8436                 struct nfs41_create_session_res *res)
8437 {
8438         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8439         /* Mark client id and session as being confirmed */
8440         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8441         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8442         session->flags = res->flags;
8443         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8444         if (res->flags & SESSION4_BACK_CHAN)
8445                 memcpy(&session->bc_attrs, &res->bc_attrs,
8446                                 sizeof(session->bc_attrs));
8447 }
8448
8449 static int _nfs4_proc_create_session(struct nfs_client *clp,
8450                 const struct cred *cred)
8451 {
8452         struct nfs4_session *session = clp->cl_session;
8453         struct nfs41_create_session_args args = {
8454                 .client = clp,
8455                 .clientid = clp->cl_clientid,
8456                 .seqid = clp->cl_seqid,
8457                 .cb_program = NFS4_CALLBACK,
8458         };
8459         struct nfs41_create_session_res res;
8460
8461         struct rpc_message msg = {
8462                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8463                 .rpc_argp = &args,
8464                 .rpc_resp = &res,
8465                 .rpc_cred = cred,
8466         };
8467         int status;
8468
8469         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8470         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8471
8472         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8473         trace_nfs4_create_session(clp, status);
8474
8475         switch (status) {
8476         case -NFS4ERR_STALE_CLIENTID:
8477         case -NFS4ERR_DELAY:
8478         case -ETIMEDOUT:
8479         case -EACCES:
8480         case -EAGAIN:
8481                 goto out;
8482         };
8483
8484         clp->cl_seqid++;
8485         if (!status) {
8486                 /* Verify the session's negotiated channel_attrs values */
8487                 status = nfs4_verify_channel_attrs(&args, &res);
8488                 /* Increment the clientid slot sequence id */
8489                 if (status)
8490                         goto out;
8491                 nfs4_update_session(session, &res);
8492         }
8493 out:
8494         return status;
8495 }
8496
8497 /*
8498  * Issues a CREATE_SESSION operation to the server.
8499  * It is the responsibility of the caller to verify the session is
8500  * expired before calling this routine.
8501  */
8502 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
8503 {
8504         int status;
8505         unsigned *ptr;
8506         struct nfs4_session *session = clp->cl_session;
8507
8508         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8509
8510         status = _nfs4_proc_create_session(clp, cred);
8511         if (status)
8512                 goto out;
8513
8514         /* Init or reset the session slot tables */
8515         status = nfs4_setup_session_slot_tables(session);
8516         dprintk("slot table setup returned %d\n", status);
8517         if (status)
8518                 goto out;
8519
8520         ptr = (unsigned *)&session->sess_id.data[0];
8521         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8522                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8523 out:
8524         dprintk("<-- %s\n", __func__);
8525         return status;
8526 }
8527
8528 /*
8529  * Issue the over-the-wire RPC DESTROY_SESSION.
8530  * The caller must serialize access to this routine.
8531  */
8532 int nfs4_proc_destroy_session(struct nfs4_session *session,
8533                 const struct cred *cred)
8534 {
8535         struct rpc_message msg = {
8536                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8537                 .rpc_argp = session,
8538                 .rpc_cred = cred,
8539         };
8540         int status = 0;
8541
8542         dprintk("--> nfs4_proc_destroy_session\n");
8543
8544         /* session is still being setup */
8545         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8546                 return 0;
8547
8548         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8549         trace_nfs4_destroy_session(session->clp, status);
8550
8551         if (status)
8552                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8553                         "Session has been destroyed regardless...\n", status);
8554
8555         dprintk("<-- nfs4_proc_destroy_session\n");
8556         return status;
8557 }
8558
8559 /*
8560  * Renew the cl_session lease.
8561  */
8562 struct nfs4_sequence_data {
8563         struct nfs_client *clp;
8564         struct nfs4_sequence_args args;
8565         struct nfs4_sequence_res res;
8566 };
8567
8568 static void nfs41_sequence_release(void *data)
8569 {
8570         struct nfs4_sequence_data *calldata = data;
8571         struct nfs_client *clp = calldata->clp;
8572
8573         if (refcount_read(&clp->cl_count) > 1)
8574                 nfs4_schedule_state_renewal(clp);
8575         nfs_put_client(clp);
8576         kfree(calldata);
8577 }
8578
8579 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8580 {
8581         switch(task->tk_status) {
8582         case -NFS4ERR_DELAY:
8583                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8584                 return -EAGAIN;
8585         default:
8586                 nfs4_schedule_lease_recovery(clp);
8587         }
8588         return 0;
8589 }
8590
8591 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8592 {
8593         struct nfs4_sequence_data *calldata = data;
8594         struct nfs_client *clp = calldata->clp;
8595
8596         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8597                 return;
8598
8599         trace_nfs4_sequence(clp, task->tk_status);
8600         if (task->tk_status < 0) {
8601                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8602                 if (refcount_read(&clp->cl_count) == 1)
8603                         goto out;
8604
8605                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8606                         rpc_restart_call_prepare(task);
8607                         return;
8608                 }
8609         }
8610         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8611 out:
8612         dprintk("<-- %s\n", __func__);
8613 }
8614
8615 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8616 {
8617         struct nfs4_sequence_data *calldata = data;
8618         struct nfs_client *clp = calldata->clp;
8619         struct nfs4_sequence_args *args;
8620         struct nfs4_sequence_res *res;
8621
8622         args = task->tk_msg.rpc_argp;
8623         res = task->tk_msg.rpc_resp;
8624
8625         nfs4_setup_sequence(clp, args, res, task);
8626 }
8627
8628 static const struct rpc_call_ops nfs41_sequence_ops = {
8629         .rpc_call_done = nfs41_sequence_call_done,
8630         .rpc_call_prepare = nfs41_sequence_prepare,
8631         .rpc_release = nfs41_sequence_release,
8632 };
8633
8634 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8635                 const struct cred *cred,
8636                 struct nfs4_slot *slot,
8637                 bool is_privileged)
8638 {
8639         struct nfs4_sequence_data *calldata;
8640         struct rpc_message msg = {
8641                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8642                 .rpc_cred = cred,
8643         };
8644         struct rpc_task_setup task_setup_data = {
8645                 .rpc_client = clp->cl_rpcclient,
8646                 .rpc_message = &msg,
8647                 .callback_ops = &nfs41_sequence_ops,
8648                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8649         };
8650         struct rpc_task *ret;
8651
8652         ret = ERR_PTR(-EIO);
8653         if (!refcount_inc_not_zero(&clp->cl_count))
8654                 goto out_err;
8655
8656         ret = ERR_PTR(-ENOMEM);
8657         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8658         if (calldata == NULL)
8659                 goto out_put_clp;
8660         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
8661         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
8662         msg.rpc_argp = &calldata->args;
8663         msg.rpc_resp = &calldata->res;
8664         calldata->clp = clp;
8665         task_setup_data.callback_data = calldata;
8666
8667         ret = rpc_run_task(&task_setup_data);
8668         if (IS_ERR(ret))
8669                 goto out_err;
8670         return ret;
8671 out_put_clp:
8672         nfs_put_client(clp);
8673 out_err:
8674         nfs41_release_slot(slot);
8675         return ret;
8676 }
8677
8678 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
8679 {
8680         struct rpc_task *task;
8681         int ret = 0;
8682
8683         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8684                 return -EAGAIN;
8685         task = _nfs41_proc_sequence(clp, cred, NULL, false);
8686         if (IS_ERR(task))
8687                 ret = PTR_ERR(task);
8688         else
8689                 rpc_put_task_async(task);
8690         dprintk("<-- %s status=%d\n", __func__, ret);
8691         return ret;
8692 }
8693
8694 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
8695 {
8696         struct rpc_task *task;
8697         int ret;
8698
8699         task = _nfs41_proc_sequence(clp, cred, NULL, true);
8700         if (IS_ERR(task)) {
8701                 ret = PTR_ERR(task);
8702                 goto out;
8703         }
8704         ret = rpc_wait_for_completion_task(task);
8705         if (!ret)
8706                 ret = task->tk_status;
8707         rpc_put_task(task);
8708 out:
8709         dprintk("<-- %s status=%d\n", __func__, ret);
8710         return ret;
8711 }
8712
8713 struct nfs4_reclaim_complete_data {
8714         struct nfs_client *clp;
8715         struct nfs41_reclaim_complete_args arg;
8716         struct nfs41_reclaim_complete_res res;
8717 };
8718
8719 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8720 {
8721         struct nfs4_reclaim_complete_data *calldata = data;
8722
8723         nfs4_setup_sequence(calldata->clp,
8724                         &calldata->arg.seq_args,
8725                         &calldata->res.seq_res,
8726                         task);
8727 }
8728
8729 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8730 {
8731         switch(task->tk_status) {
8732         case 0:
8733                 wake_up_all(&clp->cl_lock_waitq);
8734                 /* Fallthrough */
8735         case -NFS4ERR_COMPLETE_ALREADY:
8736         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8737                 break;
8738         case -NFS4ERR_DELAY:
8739                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8740                 /* fall through */
8741         case -NFS4ERR_RETRY_UNCACHED_REP:
8742                 return -EAGAIN;
8743         case -NFS4ERR_BADSESSION:
8744         case -NFS4ERR_DEADSESSION:
8745         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8746                 nfs4_schedule_session_recovery(clp->cl_session,
8747                                 task->tk_status);
8748                 break;
8749         default:
8750                 nfs4_schedule_lease_recovery(clp);
8751         }
8752         return 0;
8753 }
8754
8755 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8756 {
8757         struct nfs4_reclaim_complete_data *calldata = data;
8758         struct nfs_client *clp = calldata->clp;
8759         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8760
8761         dprintk("--> %s\n", __func__);
8762         if (!nfs41_sequence_done(task, res))
8763                 return;
8764
8765         trace_nfs4_reclaim_complete(clp, task->tk_status);
8766         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8767                 rpc_restart_call_prepare(task);
8768                 return;
8769         }
8770         dprintk("<-- %s\n", __func__);
8771 }
8772
8773 static void nfs4_free_reclaim_complete_data(void *data)
8774 {
8775         struct nfs4_reclaim_complete_data *calldata = data;
8776
8777         kfree(calldata);
8778 }
8779
8780 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8781         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8782         .rpc_call_done = nfs4_reclaim_complete_done,
8783         .rpc_release = nfs4_free_reclaim_complete_data,
8784 };
8785
8786 /*
8787  * Issue a global reclaim complete.
8788  */
8789 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8790                 const struct cred *cred)
8791 {
8792         struct nfs4_reclaim_complete_data *calldata;
8793         struct rpc_task *task;
8794         struct rpc_message msg = {
8795                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8796                 .rpc_cred = cred,
8797         };
8798         struct rpc_task_setup task_setup_data = {
8799                 .rpc_client = clp->cl_rpcclient,
8800                 .rpc_message = &msg,
8801                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8802                 .flags = RPC_TASK_ASYNC,
8803         };
8804         int status = -ENOMEM;
8805
8806         dprintk("--> %s\n", __func__);
8807         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8808         if (calldata == NULL)
8809                 goto out;
8810         calldata->clp = clp;
8811         calldata->arg.one_fs = 0;
8812
8813         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
8814         msg.rpc_argp = &calldata->arg;
8815         msg.rpc_resp = &calldata->res;
8816         task_setup_data.callback_data = calldata;
8817         task = rpc_run_task(&task_setup_data);
8818         if (IS_ERR(task)) {
8819                 status = PTR_ERR(task);
8820                 goto out;
8821         }
8822         status = rpc_wait_for_completion_task(task);
8823         if (status == 0)
8824                 status = task->tk_status;
8825         rpc_put_task(task);
8826 out:
8827         dprintk("<-- %s status=%d\n", __func__, status);
8828         return status;
8829 }
8830
8831 static void
8832 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8833 {
8834         struct nfs4_layoutget *lgp = calldata;
8835         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8836
8837         dprintk("--> %s\n", __func__);
8838         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
8839                                 &lgp->res.seq_res, task);
8840         dprintk("<-- %s\n", __func__);
8841 }
8842
8843 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8844 {
8845         struct nfs4_layoutget *lgp = calldata;
8846
8847         dprintk("--> %s\n", __func__);
8848         nfs41_sequence_process(task, &lgp->res.seq_res);
8849         dprintk("<-- %s\n", __func__);
8850 }
8851
8852 static int
8853 nfs4_layoutget_handle_exception(struct rpc_task *task,
8854                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8855 {
8856         struct inode *inode = lgp->args.inode;
8857         struct nfs_server *server = NFS_SERVER(inode);
8858         struct pnfs_layout_hdr *lo;
8859         int nfs4err = task->tk_status;
8860         int err, status = 0;
8861         LIST_HEAD(head);
8862
8863         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8864
8865         nfs4_sequence_free_slot(&lgp->res.seq_res);
8866
8867         switch (nfs4err) {
8868         case 0:
8869                 goto out;
8870
8871         /*
8872          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8873          * on the file. set tk_status to -ENODATA to tell upper layer to
8874          * retry go inband.
8875          */
8876         case -NFS4ERR_LAYOUTUNAVAILABLE:
8877                 status = -ENODATA;
8878                 goto out;
8879         /*
8880          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8881          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8882          */
8883         case -NFS4ERR_BADLAYOUT:
8884                 status = -EOVERFLOW;
8885                 goto out;
8886         /*
8887          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8888          * (or clients) writing to the same RAID stripe except when
8889          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8890          *
8891          * Treat it like we would RECALLCONFLICT -- we retry for a little
8892          * while, and then eventually give up.
8893          */
8894         case -NFS4ERR_LAYOUTTRYLATER:
8895                 if (lgp->args.minlength == 0) {
8896                         status = -EOVERFLOW;
8897                         goto out;
8898                 }
8899                 status = -EBUSY;
8900                 break;
8901         case -NFS4ERR_RECALLCONFLICT:
8902                 status = -ERECALLCONFLICT;
8903                 break;
8904         case -NFS4ERR_DELEG_REVOKED:
8905         case -NFS4ERR_ADMIN_REVOKED:
8906         case -NFS4ERR_EXPIRED:
8907         case -NFS4ERR_BAD_STATEID:
8908                 exception->timeout = 0;
8909                 spin_lock(&inode->i_lock);
8910                 lo = NFS_I(inode)->layout;
8911                 /* If the open stateid was bad, then recover it. */
8912                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8913                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
8914                         spin_unlock(&inode->i_lock);
8915                         exception->state = lgp->args.ctx->state;
8916                         exception->stateid = &lgp->args.stateid;
8917                         break;
8918                 }
8919
8920                 /*
8921                  * Mark the bad layout state as invalid, then retry
8922                  */
8923                 pnfs_mark_layout_stateid_invalid(lo, &head);
8924                 spin_unlock(&inode->i_lock);
8925                 nfs_commit_inode(inode, 0);
8926                 pnfs_free_lseg_list(&head);
8927                 status = -EAGAIN;
8928                 goto out;
8929         }
8930
8931         err = nfs4_handle_exception(server, nfs4err, exception);
8932         if (!status) {
8933                 if (exception->retry)
8934                         status = -EAGAIN;
8935                 else
8936                         status = err;
8937         }
8938 out:
8939         dprintk("<-- %s\n", __func__);
8940         return status;
8941 }
8942
8943 size_t max_response_pages(struct nfs_server *server)
8944 {
8945         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8946         return nfs_page_array_len(0, max_resp_sz);
8947 }
8948
8949 static void nfs4_layoutget_release(void *calldata)
8950 {
8951         struct nfs4_layoutget *lgp = calldata;
8952
8953         dprintk("--> %s\n", __func__);
8954         nfs4_sequence_free_slot(&lgp->res.seq_res);
8955         pnfs_layoutget_free(lgp);
8956         dprintk("<-- %s\n", __func__);
8957 }
8958
8959 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8960         .rpc_call_prepare = nfs4_layoutget_prepare,
8961         .rpc_call_done = nfs4_layoutget_done,
8962         .rpc_release = nfs4_layoutget_release,
8963 };
8964
8965 struct pnfs_layout_segment *
8966 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
8967 {
8968         struct inode *inode = lgp->args.inode;
8969         struct nfs_server *server = NFS_SERVER(inode);
8970         struct rpc_task *task;
8971         struct rpc_message msg = {
8972                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8973                 .rpc_argp = &lgp->args,
8974                 .rpc_resp = &lgp->res,
8975                 .rpc_cred = lgp->cred,
8976         };
8977         struct rpc_task_setup task_setup_data = {
8978                 .rpc_client = server->client,
8979                 .rpc_message = &msg,
8980                 .callback_ops = &nfs4_layoutget_call_ops,
8981                 .callback_data = lgp,
8982                 .flags = RPC_TASK_ASYNC,
8983         };
8984         struct pnfs_layout_segment *lseg = NULL;
8985         struct nfs4_exception exception = {
8986                 .inode = inode,
8987                 .timeout = *timeout,
8988         };
8989         int status = 0;
8990
8991         dprintk("--> %s\n", __func__);
8992
8993         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8994         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8995
8996         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
8997
8998         task = rpc_run_task(&task_setup_data);
8999         if (IS_ERR(task))
9000                 return ERR_CAST(task);
9001         status = rpc_wait_for_completion_task(task);
9002         if (status != 0)
9003                 goto out;
9004
9005         if (task->tk_status < 0) {
9006                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9007                 *timeout = exception.timeout;
9008         } else if (lgp->res.layoutp->len == 0) {
9009                 status = -EAGAIN;
9010                 *timeout = nfs4_update_delay(&exception.timeout);
9011         } else
9012                 lseg = pnfs_layout_process(lgp);
9013 out:
9014         trace_nfs4_layoutget(lgp->args.ctx,
9015                         &lgp->args.range,
9016                         &lgp->res.range,
9017                         &lgp->res.stateid,
9018                         status);
9019
9020         rpc_put_task(task);
9021         dprintk("<-- %s status=%d\n", __func__, status);
9022         if (status)
9023                 return ERR_PTR(status);
9024         return lseg;
9025 }
9026
9027 static void
9028 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9029 {
9030         struct nfs4_layoutreturn *lrp = calldata;
9031
9032         dprintk("--> %s\n", __func__);
9033         nfs4_setup_sequence(lrp->clp,
9034                         &lrp->args.seq_args,
9035                         &lrp->res.seq_res,
9036                         task);
9037         if (!pnfs_layout_is_valid(lrp->args.layout))
9038                 rpc_exit(task, 0);
9039 }
9040
9041 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9042 {
9043         struct nfs4_layoutreturn *lrp = calldata;
9044         struct nfs_server *server;
9045
9046         dprintk("--> %s\n", __func__);
9047
9048         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9049                 return;
9050
9051         server = NFS_SERVER(lrp->args.inode);
9052         switch (task->tk_status) {
9053         case -NFS4ERR_OLD_STATEID:
9054                 if (nfs4_layoutreturn_refresh_stateid(&lrp->args.stateid,
9055                                         &lrp->args.range,
9056                                         lrp->args.inode))
9057                         goto out_restart;
9058                 /* Fallthrough */
9059         default:
9060                 task->tk_status = 0;
9061                 /* Fallthrough */
9062         case 0:
9063                 break;
9064         case -NFS4ERR_DELAY:
9065                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9066                         break;
9067                 goto out_restart;
9068         }
9069         dprintk("<-- %s\n", __func__);
9070         return;
9071 out_restart:
9072         task->tk_status = 0;
9073         nfs4_sequence_free_slot(&lrp->res.seq_res);
9074         rpc_restart_call_prepare(task);
9075 }
9076
9077 static void nfs4_layoutreturn_release(void *calldata)
9078 {
9079         struct nfs4_layoutreturn *lrp = calldata;
9080         struct pnfs_layout_hdr *lo = lrp->args.layout;
9081
9082         dprintk("--> %s\n", __func__);
9083         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9084                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9085         nfs4_sequence_free_slot(&lrp->res.seq_res);
9086         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9087                 lrp->ld_private.ops->free(&lrp->ld_private);
9088         pnfs_put_layout_hdr(lrp->args.layout);
9089         nfs_iput_and_deactive(lrp->inode);
9090         kfree(calldata);
9091         dprintk("<-- %s\n", __func__);
9092 }
9093
9094 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9095         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9096         .rpc_call_done = nfs4_layoutreturn_done,
9097         .rpc_release = nfs4_layoutreturn_release,
9098 };
9099
9100 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9101 {
9102         struct rpc_task *task;
9103         struct rpc_message msg = {
9104                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9105                 .rpc_argp = &lrp->args,
9106                 .rpc_resp = &lrp->res,
9107                 .rpc_cred = lrp->cred,
9108         };
9109         struct rpc_task_setup task_setup_data = {
9110                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9111                 .rpc_message = &msg,
9112                 .callback_ops = &nfs4_layoutreturn_call_ops,
9113                 .callback_data = lrp,
9114         };
9115         int status = 0;
9116
9117         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9118                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9119                         &task_setup_data.rpc_client, &msg);
9120
9121         dprintk("--> %s\n", __func__);
9122         if (!sync) {
9123                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9124                 if (!lrp->inode) {
9125                         nfs4_layoutreturn_release(lrp);
9126                         return -EAGAIN;
9127                 }
9128                 task_setup_data.flags |= RPC_TASK_ASYNC;
9129         }
9130         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9131         task = rpc_run_task(&task_setup_data);
9132         if (IS_ERR(task))
9133                 return PTR_ERR(task);
9134         if (sync)
9135                 status = task->tk_status;
9136         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9137         dprintk("<-- %s status=%d\n", __func__, status);
9138         rpc_put_task(task);
9139         return status;
9140 }
9141
9142 static int
9143 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9144                 struct pnfs_device *pdev,
9145                 const struct cred *cred)
9146 {
9147         struct nfs4_getdeviceinfo_args args = {
9148                 .pdev = pdev,
9149                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9150                         NOTIFY_DEVICEID4_DELETE,
9151         };
9152         struct nfs4_getdeviceinfo_res res = {
9153                 .pdev = pdev,
9154         };
9155         struct rpc_message msg = {
9156                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9157                 .rpc_argp = &args,
9158                 .rpc_resp = &res,
9159                 .rpc_cred = cred,
9160         };
9161         int status;
9162
9163         dprintk("--> %s\n", __func__);
9164         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9165         if (res.notification & ~args.notify_types)
9166                 dprintk("%s: unsupported notification\n", __func__);
9167         if (res.notification != args.notify_types)
9168                 pdev->nocache = 1;
9169
9170         dprintk("<-- %s status=%d\n", __func__, status);
9171
9172         return status;
9173 }
9174
9175 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9176                 struct pnfs_device *pdev,
9177                 const struct cred *cred)
9178 {
9179         struct nfs4_exception exception = { };
9180         int err;
9181
9182         do {
9183                 err = nfs4_handle_exception(server,
9184                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9185                                         &exception);
9186         } while (exception.retry);
9187         return err;
9188 }
9189 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9190
9191 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9192 {
9193         struct nfs4_layoutcommit_data *data = calldata;
9194         struct nfs_server *server = NFS_SERVER(data->args.inode);
9195
9196         nfs4_setup_sequence(server->nfs_client,
9197                         &data->args.seq_args,
9198                         &data->res.seq_res,
9199                         task);
9200 }
9201
9202 static void
9203 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9204 {
9205         struct nfs4_layoutcommit_data *data = calldata;
9206         struct nfs_server *server = NFS_SERVER(data->args.inode);
9207
9208         if (!nfs41_sequence_done(task, &data->res.seq_res))
9209                 return;
9210
9211         switch (task->tk_status) { /* Just ignore these failures */
9212         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9213         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9214         case -NFS4ERR_BADLAYOUT:     /* no layout */
9215         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9216                 task->tk_status = 0;
9217         case 0:
9218                 break;
9219         default:
9220                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9221                         rpc_restart_call_prepare(task);
9222                         return;
9223                 }
9224         }
9225 }
9226
9227 static void nfs4_layoutcommit_release(void *calldata)
9228 {
9229         struct nfs4_layoutcommit_data *data = calldata;
9230
9231         pnfs_cleanup_layoutcommit(data);
9232         nfs_post_op_update_inode_force_wcc(data->args.inode,
9233                                            data->res.fattr);
9234         put_cred(data->cred);
9235         nfs_iput_and_deactive(data->inode);
9236         kfree(data);
9237 }
9238
9239 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9240         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9241         .rpc_call_done = nfs4_layoutcommit_done,
9242         .rpc_release = nfs4_layoutcommit_release,
9243 };
9244
9245 int
9246 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9247 {
9248         struct rpc_message msg = {
9249                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9250                 .rpc_argp = &data->args,
9251                 .rpc_resp = &data->res,
9252                 .rpc_cred = data->cred,
9253         };
9254         struct rpc_task_setup task_setup_data = {
9255                 .task = &data->task,
9256                 .rpc_client = NFS_CLIENT(data->args.inode),
9257                 .rpc_message = &msg,
9258                 .callback_ops = &nfs4_layoutcommit_ops,
9259                 .callback_data = data,
9260         };
9261         struct rpc_task *task;
9262         int status = 0;
9263
9264         dprintk("NFS: initiating layoutcommit call. sync %d "
9265                 "lbw: %llu inode %lu\n", sync,
9266                 data->args.lastbytewritten,
9267                 data->args.inode->i_ino);
9268
9269         if (!sync) {
9270                 data->inode = nfs_igrab_and_active(data->args.inode);
9271                 if (data->inode == NULL) {
9272                         nfs4_layoutcommit_release(data);
9273                         return -EAGAIN;
9274                 }
9275                 task_setup_data.flags = RPC_TASK_ASYNC;
9276         }
9277         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9278         task = rpc_run_task(&task_setup_data);
9279         if (IS_ERR(task))
9280                 return PTR_ERR(task);
9281         if (sync)
9282                 status = task->tk_status;
9283         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9284         dprintk("%s: status %d\n", __func__, status);
9285         rpc_put_task(task);
9286         return status;
9287 }
9288
9289 /*
9290  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9291  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9292  */
9293 static int
9294 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9295                     struct nfs_fsinfo *info,
9296                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9297 {
9298         struct nfs41_secinfo_no_name_args args = {
9299                 .style = SECINFO_STYLE_CURRENT_FH,
9300         };
9301         struct nfs4_secinfo_res res = {
9302                 .flavors = flavors,
9303         };
9304         struct rpc_message msg = {
9305                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9306                 .rpc_argp = &args,
9307                 .rpc_resp = &res,
9308         };
9309         struct rpc_clnt *clnt = server->client;
9310         const struct cred *cred = NULL;
9311         int status;
9312
9313         if (use_integrity) {
9314                 clnt = server->nfs_client->cl_rpcclient;
9315                 cred = nfs4_get_clid_cred(server->nfs_client);
9316                 msg.rpc_cred = cred;
9317         }
9318
9319         dprintk("--> %s\n", __func__);
9320         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
9321                                 &res.seq_res, 0);
9322         dprintk("<-- %s status=%d\n", __func__, status);
9323
9324         put_cred(cred);
9325
9326         return status;
9327 }
9328
9329 static int
9330 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9331                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9332 {
9333         struct nfs4_exception exception = {
9334                 .interruptible = true,
9335         };
9336         int err;
9337         do {
9338                 /* first try using integrity protection */
9339                 err = -NFS4ERR_WRONGSEC;
9340
9341                 /* try to use integrity protection with machine cred */
9342                 if (_nfs4_is_integrity_protected(server->nfs_client))
9343                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9344                                                           flavors, true);
9345
9346                 /*
9347                  * if unable to use integrity protection, or SECINFO with
9348                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9349                  * disallowed by spec, but exists in deployed servers) use
9350                  * the current filesystem's rpc_client and the user cred.
9351                  */
9352                 if (err == -NFS4ERR_WRONGSEC)
9353                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9354                                                           flavors, false);
9355
9356                 switch (err) {
9357                 case 0:
9358                 case -NFS4ERR_WRONGSEC:
9359                 case -ENOTSUPP:
9360                         goto out;
9361                 default:
9362                         err = nfs4_handle_exception(server, err, &exception);
9363                 }
9364         } while (exception.retry);
9365 out:
9366         return err;
9367 }
9368
9369 static int
9370 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9371                     struct nfs_fsinfo *info)
9372 {
9373         int err;
9374         struct page *page;
9375         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9376         struct nfs4_secinfo_flavors *flavors;
9377         struct nfs4_secinfo4 *secinfo;
9378         int i;
9379
9380         page = alloc_page(GFP_KERNEL);
9381         if (!page) {
9382                 err = -ENOMEM;
9383                 goto out;
9384         }
9385
9386         flavors = page_address(page);
9387         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9388
9389         /*
9390          * Fall back on "guess and check" method if
9391          * the server doesn't support SECINFO_NO_NAME
9392          */
9393         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9394                 err = nfs4_find_root_sec(server, fhandle, info);
9395                 goto out_freepage;
9396         }
9397         if (err)
9398                 goto out_freepage;
9399
9400         for (i = 0; i < flavors->num_flavors; i++) {
9401                 secinfo = &flavors->flavors[i];
9402
9403                 switch (secinfo->flavor) {
9404                 case RPC_AUTH_NULL:
9405                 case RPC_AUTH_UNIX:
9406                 case RPC_AUTH_GSS:
9407                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9408                                         &secinfo->flavor_info);
9409                         break;
9410                 default:
9411                         flavor = RPC_AUTH_MAXFLAVOR;
9412                         break;
9413                 }
9414
9415                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9416                         flavor = RPC_AUTH_MAXFLAVOR;
9417
9418                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9419                         err = nfs4_lookup_root_sec(server, fhandle,
9420                                                    info, flavor);
9421                         if (!err)
9422                                 break;
9423                 }
9424         }
9425
9426         if (flavor == RPC_AUTH_MAXFLAVOR)
9427                 err = -EPERM;
9428
9429 out_freepage:
9430         put_page(page);
9431         if (err == -EACCES)
9432                 return -EPERM;
9433 out:
9434         return err;
9435 }
9436
9437 static int _nfs41_test_stateid(struct nfs_server *server,
9438                 nfs4_stateid *stateid,
9439                 const struct cred *cred)
9440 {
9441         int status;
9442         struct nfs41_test_stateid_args args = {
9443                 .stateid = stateid,
9444         };
9445         struct nfs41_test_stateid_res res;
9446         struct rpc_message msg = {
9447                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9448                 .rpc_argp = &args,
9449                 .rpc_resp = &res,
9450                 .rpc_cred = cred,
9451         };
9452         struct rpc_clnt *rpc_client = server->client;
9453
9454         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9455                 &rpc_client, &msg);
9456
9457         dprintk("NFS call  test_stateid %p\n", stateid);
9458         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9459         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9460                         &args.seq_args, &res.seq_res);
9461         if (status != NFS_OK) {
9462                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9463                 return status;
9464         }
9465         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9466         return -res.status;
9467 }
9468
9469 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9470                 int err, struct nfs4_exception *exception)
9471 {
9472         exception->retry = 0;
9473         switch(err) {
9474         case -NFS4ERR_DELAY:
9475         case -NFS4ERR_RETRY_UNCACHED_REP:
9476                 nfs4_handle_exception(server, err, exception);
9477                 break;
9478         case -NFS4ERR_BADSESSION:
9479         case -NFS4ERR_BADSLOT:
9480         case -NFS4ERR_BAD_HIGH_SLOT:
9481         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9482         case -NFS4ERR_DEADSESSION:
9483                 nfs4_do_handle_exception(server, err, exception);
9484         }
9485 }
9486
9487 /**
9488  * nfs41_test_stateid - perform a TEST_STATEID operation
9489  *
9490  * @server: server / transport on which to perform the operation
9491  * @stateid: state ID to test
9492  * @cred: credential
9493  *
9494  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9495  * Otherwise a negative NFS4ERR value is returned if the operation
9496  * failed or the state ID is not currently valid.
9497  */
9498 static int nfs41_test_stateid(struct nfs_server *server,
9499                 nfs4_stateid *stateid,
9500                 const struct cred *cred)
9501 {
9502         struct nfs4_exception exception = {
9503                 .interruptible = true,
9504         };
9505         int err;
9506         do {
9507                 err = _nfs41_test_stateid(server, stateid, cred);
9508                 nfs4_handle_delay_or_session_error(server, err, &exception);
9509         } while (exception.retry);
9510         return err;
9511 }
9512
9513 struct nfs_free_stateid_data {
9514         struct nfs_server *server;
9515         struct nfs41_free_stateid_args args;
9516         struct nfs41_free_stateid_res res;
9517 };
9518
9519 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9520 {
9521         struct nfs_free_stateid_data *data = calldata;
9522         nfs4_setup_sequence(data->server->nfs_client,
9523                         &data->args.seq_args,
9524                         &data->res.seq_res,
9525                         task);
9526 }
9527
9528 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9529 {
9530         struct nfs_free_stateid_data *data = calldata;
9531
9532         nfs41_sequence_done(task, &data->res.seq_res);
9533
9534         switch (task->tk_status) {
9535         case -NFS4ERR_DELAY:
9536                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9537                         rpc_restart_call_prepare(task);
9538         }
9539 }
9540
9541 static void nfs41_free_stateid_release(void *calldata)
9542 {
9543         kfree(calldata);
9544 }
9545
9546 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9547         .rpc_call_prepare = nfs41_free_stateid_prepare,
9548         .rpc_call_done = nfs41_free_stateid_done,
9549         .rpc_release = nfs41_free_stateid_release,
9550 };
9551
9552 /**
9553  * nfs41_free_stateid - perform a FREE_STATEID operation
9554  *
9555  * @server: server / transport on which to perform the operation
9556  * @stateid: state ID to release
9557  * @cred: credential
9558  * @privileged: set to true if this call needs to be privileged
9559  *
9560  * Note: this function is always asynchronous.
9561  */
9562 static int nfs41_free_stateid(struct nfs_server *server,
9563                 const nfs4_stateid *stateid,
9564                 const struct cred *cred,
9565                 bool privileged)
9566 {
9567         struct rpc_message msg = {
9568                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9569                 .rpc_cred = cred,
9570         };
9571         struct rpc_task_setup task_setup = {
9572                 .rpc_client = server->client,
9573                 .rpc_message = &msg,
9574                 .callback_ops = &nfs41_free_stateid_ops,
9575                 .flags = RPC_TASK_ASYNC,
9576         };
9577         struct nfs_free_stateid_data *data;
9578         struct rpc_task *task;
9579
9580         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9581                 &task_setup.rpc_client, &msg);
9582
9583         dprintk("NFS call  free_stateid %p\n", stateid);
9584         data = kmalloc(sizeof(*data), GFP_NOFS);
9585         if (!data)
9586                 return -ENOMEM;
9587         data->server = server;
9588         nfs4_stateid_copy(&data->args.stateid, stateid);
9589
9590         task_setup.callback_data = data;
9591
9592         msg.rpc_argp = &data->args;
9593         msg.rpc_resp = &data->res;
9594         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
9595         task = rpc_run_task(&task_setup);
9596         if (IS_ERR(task))
9597                 return PTR_ERR(task);
9598         rpc_put_task(task);
9599         return 0;
9600 }
9601
9602 static void
9603 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9604 {
9605         const struct cred *cred = lsp->ls_state->owner->so_cred;
9606
9607         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9608         nfs4_free_lock_state(server, lsp);
9609 }
9610
9611 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9612                 const nfs4_stateid *s2)
9613 {
9614         if (s1->type != s2->type)
9615                 return false;
9616
9617         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9618                 return false;
9619
9620         if (s1->seqid == s2->seqid)
9621                 return true;
9622
9623         return s1->seqid == 0 || s2->seqid == 0;
9624 }
9625
9626 #endif /* CONFIG_NFS_V4_1 */
9627
9628 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9629                 const nfs4_stateid *s2)
9630 {
9631         return nfs4_stateid_match(s1, s2);
9632 }
9633
9634
9635 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9636         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9637         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9638         .recover_open   = nfs4_open_reclaim,
9639         .recover_lock   = nfs4_lock_reclaim,
9640         .establish_clid = nfs4_init_clientid,
9641         .detect_trunking = nfs40_discover_server_trunking,
9642 };
9643
9644 #if defined(CONFIG_NFS_V4_1)
9645 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9646         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9647         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9648         .recover_open   = nfs4_open_reclaim,
9649         .recover_lock   = nfs4_lock_reclaim,
9650         .establish_clid = nfs41_init_clientid,
9651         .reclaim_complete = nfs41_proc_reclaim_complete,
9652         .detect_trunking = nfs41_discover_server_trunking,
9653 };
9654 #endif /* CONFIG_NFS_V4_1 */
9655
9656 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9657         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9658         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9659         .recover_open   = nfs40_open_expired,
9660         .recover_lock   = nfs4_lock_expired,
9661         .establish_clid = nfs4_init_clientid,
9662 };
9663
9664 #if defined(CONFIG_NFS_V4_1)
9665 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9666         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9667         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9668         .recover_open   = nfs41_open_expired,
9669         .recover_lock   = nfs41_lock_expired,
9670         .establish_clid = nfs41_init_clientid,
9671 };
9672 #endif /* CONFIG_NFS_V4_1 */
9673
9674 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9675         .sched_state_renewal = nfs4_proc_async_renew,
9676         .get_state_renewal_cred = nfs4_get_renew_cred,
9677         .renew_lease = nfs4_proc_renew,
9678 };
9679
9680 #if defined(CONFIG_NFS_V4_1)
9681 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9682         .sched_state_renewal = nfs41_proc_async_sequence,
9683         .get_state_renewal_cred = nfs4_get_machine_cred,
9684         .renew_lease = nfs4_proc_sequence,
9685 };
9686 #endif
9687
9688 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9689         .get_locations = _nfs40_proc_get_locations,
9690         .fsid_present = _nfs40_proc_fsid_present,
9691 };
9692
9693 #if defined(CONFIG_NFS_V4_1)
9694 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9695         .get_locations = _nfs41_proc_get_locations,
9696         .fsid_present = _nfs41_proc_fsid_present,
9697 };
9698 #endif  /* CONFIG_NFS_V4_1 */
9699
9700 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9701         .minor_version = 0,
9702         .init_caps = NFS_CAP_READDIRPLUS
9703                 | NFS_CAP_ATOMIC_OPEN
9704                 | NFS_CAP_POSIX_LOCK,
9705         .init_client = nfs40_init_client,
9706         .shutdown_client = nfs40_shutdown_client,
9707         .match_stateid = nfs4_match_stateid,
9708         .find_root_sec = nfs4_find_root_sec,
9709         .free_lock_state = nfs4_release_lockowner,
9710         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9711         .alloc_seqid = nfs_alloc_seqid,
9712         .call_sync_ops = &nfs40_call_sync_ops,
9713         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9714         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9715         .state_renewal_ops = &nfs40_state_renewal_ops,
9716         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9717 };
9718
9719 #if defined(CONFIG_NFS_V4_1)
9720 static struct nfs_seqid *
9721 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9722 {
9723         return NULL;
9724 }
9725
9726 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9727         .minor_version = 1,
9728         .init_caps = NFS_CAP_READDIRPLUS
9729                 | NFS_CAP_ATOMIC_OPEN
9730                 | NFS_CAP_POSIX_LOCK
9731                 | NFS_CAP_STATEID_NFSV41
9732                 | NFS_CAP_ATOMIC_OPEN_V1
9733                 | NFS_CAP_LGOPEN,
9734         .init_client = nfs41_init_client,
9735         .shutdown_client = nfs41_shutdown_client,
9736         .match_stateid = nfs41_match_stateid,
9737         .find_root_sec = nfs41_find_root_sec,
9738         .free_lock_state = nfs41_free_lock_state,
9739         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9740         .alloc_seqid = nfs_alloc_no_seqid,
9741         .session_trunk = nfs4_test_session_trunk,
9742         .call_sync_ops = &nfs41_call_sync_ops,
9743         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9744         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9745         .state_renewal_ops = &nfs41_state_renewal_ops,
9746         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9747 };
9748 #endif
9749
9750 #if defined(CONFIG_NFS_V4_2)
9751 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9752         .minor_version = 2,
9753         .init_caps = NFS_CAP_READDIRPLUS
9754                 | NFS_CAP_ATOMIC_OPEN
9755                 | NFS_CAP_POSIX_LOCK
9756                 | NFS_CAP_STATEID_NFSV41
9757                 | NFS_CAP_ATOMIC_OPEN_V1
9758                 | NFS_CAP_LGOPEN
9759                 | NFS_CAP_ALLOCATE
9760                 | NFS_CAP_COPY
9761                 | NFS_CAP_OFFLOAD_CANCEL
9762                 | NFS_CAP_DEALLOCATE
9763                 | NFS_CAP_SEEK
9764                 | NFS_CAP_LAYOUTSTATS
9765                 | NFS_CAP_CLONE
9766                 | NFS_CAP_LAYOUTERROR,
9767         .init_client = nfs41_init_client,
9768         .shutdown_client = nfs41_shutdown_client,
9769         .match_stateid = nfs41_match_stateid,
9770         .find_root_sec = nfs41_find_root_sec,
9771         .free_lock_state = nfs41_free_lock_state,
9772         .call_sync_ops = &nfs41_call_sync_ops,
9773         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9774         .alloc_seqid = nfs_alloc_no_seqid,
9775         .session_trunk = nfs4_test_session_trunk,
9776         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9777         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9778         .state_renewal_ops = &nfs41_state_renewal_ops,
9779         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9780 };
9781 #endif
9782
9783 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9784         [0] = &nfs_v4_0_minor_ops,
9785 #if defined(CONFIG_NFS_V4_1)
9786         [1] = &nfs_v4_1_minor_ops,
9787 #endif
9788 #if defined(CONFIG_NFS_V4_2)
9789         [2] = &nfs_v4_2_minor_ops,
9790 #endif
9791 };
9792
9793 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9794 {
9795         ssize_t error, error2;
9796
9797         error = generic_listxattr(dentry, list, size);
9798         if (error < 0)
9799                 return error;
9800         if (list) {
9801                 list += error;
9802                 size -= error;
9803         }
9804
9805         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9806         if (error2 < 0)
9807                 return error2;
9808         return error + error2;
9809 }
9810
9811 static const struct inode_operations nfs4_dir_inode_operations = {
9812         .create         = nfs_create,
9813         .lookup         = nfs_lookup,
9814         .atomic_open    = nfs_atomic_open,
9815         .link           = nfs_link,
9816         .unlink         = nfs_unlink,
9817         .symlink        = nfs_symlink,
9818         .mkdir          = nfs_mkdir,
9819         .rmdir          = nfs_rmdir,
9820         .mknod          = nfs_mknod,
9821         .rename         = nfs_rename,
9822         .permission     = nfs_permission,
9823         .getattr        = nfs_getattr,
9824         .setattr        = nfs_setattr,
9825         .listxattr      = nfs4_listxattr,
9826 };
9827
9828 static const struct inode_operations nfs4_file_inode_operations = {
9829         .permission     = nfs_permission,
9830         .getattr        = nfs_getattr,
9831         .setattr        = nfs_setattr,
9832         .listxattr      = nfs4_listxattr,
9833 };
9834
9835 const struct nfs_rpc_ops nfs_v4_clientops = {
9836         .version        = 4,                    /* protocol version */
9837         .dentry_ops     = &nfs4_dentry_operations,
9838         .dir_inode_ops  = &nfs4_dir_inode_operations,
9839         .file_inode_ops = &nfs4_file_inode_operations,
9840         .file_ops       = &nfs4_file_operations,
9841         .getroot        = nfs4_proc_get_root,
9842         .submount       = nfs4_submount,
9843         .try_mount      = nfs4_try_mount,
9844         .getattr        = nfs4_proc_getattr,
9845         .setattr        = nfs4_proc_setattr,
9846         .lookup         = nfs4_proc_lookup,
9847         .lookupp        = nfs4_proc_lookupp,
9848         .access         = nfs4_proc_access,
9849         .readlink       = nfs4_proc_readlink,
9850         .create         = nfs4_proc_create,
9851         .remove         = nfs4_proc_remove,
9852         .unlink_setup   = nfs4_proc_unlink_setup,
9853         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9854         .unlink_done    = nfs4_proc_unlink_done,
9855         .rename_setup   = nfs4_proc_rename_setup,
9856         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9857         .rename_done    = nfs4_proc_rename_done,
9858         .link           = nfs4_proc_link,
9859         .symlink        = nfs4_proc_symlink,
9860         .mkdir          = nfs4_proc_mkdir,
9861         .rmdir          = nfs4_proc_rmdir,
9862         .readdir        = nfs4_proc_readdir,
9863         .mknod          = nfs4_proc_mknod,
9864         .statfs         = nfs4_proc_statfs,
9865         .fsinfo         = nfs4_proc_fsinfo,
9866         .pathconf       = nfs4_proc_pathconf,
9867         .set_capabilities = nfs4_server_capabilities,
9868         .decode_dirent  = nfs4_decode_dirent,
9869         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9870         .read_setup     = nfs4_proc_read_setup,
9871         .read_done      = nfs4_read_done,
9872         .write_setup    = nfs4_proc_write_setup,
9873         .write_done     = nfs4_write_done,
9874         .commit_setup   = nfs4_proc_commit_setup,
9875         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9876         .commit_done    = nfs4_commit_done,
9877         .lock           = nfs4_proc_lock,
9878         .clear_acl_cache = nfs4_zap_acl_attr,
9879         .close_context  = nfs4_close_context,
9880         .open_context   = nfs4_atomic_open,
9881         .have_delegation = nfs4_have_delegation,
9882         .alloc_client   = nfs4_alloc_client,
9883         .init_client    = nfs4_init_client,
9884         .free_client    = nfs4_free_client,
9885         .create_server  = nfs4_create_server,
9886         .clone_server   = nfs_clone_server,
9887 };
9888
9889 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9890         .name   = XATTR_NAME_NFSV4_ACL,
9891         .list   = nfs4_xattr_list_nfs4_acl,
9892         .get    = nfs4_xattr_get_nfs4_acl,
9893         .set    = nfs4_xattr_set_nfs4_acl,
9894 };
9895
9896 const struct xattr_handler *nfs4_xattr_handlers[] = {
9897         &nfs4_xattr_nfs4_acl_handler,
9898 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9899         &nfs4_xattr_nfs4_label_handler,
9900 #endif
9901         NULL
9902 };
9903
9904 /*
9905  * Local variables:
9906  *  c-basic-offset: 8
9907  * End:
9908  */